CA2089478C - Automatic length-reducing audio delay line - Google Patents
Automatic length-reducing audio delay lineInfo
- Publication number
- CA2089478C CA2089478C CA002089478A CA2089478A CA2089478C CA 2089478 C CA2089478 C CA 2089478C CA 002089478 A CA002089478 A CA 002089478A CA 2089478 A CA2089478 A CA 2089478A CA 2089478 C CA2089478 C CA 2089478C
- Authority
- CA
- Canada
- Prior art keywords
- buffer
- input signal
- length
- input
- output
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Processing of the speech or voice signal to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/04—Time compression or expansion
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M2250/00—Details of telephonic subscriber devices
- H04M2250/74—Details of telephonic subscriber devices with voice recognition means
Abstract
An electronic system and a method substantially eliminates any delays caused by buffering of a signal containing speech.
The electronic system comprises a receiver for receiving signals (16), a buffer (26), coupled to the receiver, a buffer controller (28) coupled to the buffer, and an audio section (32) for presenting signals received to a listener. The method comprises the following steps. The receiver receives an input signal representing speech, and produces an output signal representing the input signal after a delay intentionally introduced in the system. The delay corresponds to the length of the buffer. The buffer stores the input sig-nal in the buffer at an input rate, and produces the output signal at an output rate. The buffer controller determines whether the delay or buffer length is greater than a predetermined threshold, and analyzes at least a portion of the input signal stored in the buffer to determine which parts of the input signal may be removed without substantially deterioration of the signal quality accord-ing to at least one predetermined criterion, when the buffer length is greater than a predetermined length. The buffer controller then deletes at least some of the parts of the input signal stored in the buffer that comply with the predetermined criterion.
The electronic system comprises a receiver for receiving signals (16), a buffer (26), coupled to the receiver, a buffer controller (28) coupled to the buffer, and an audio section (32) for presenting signals received to a listener. The method comprises the following steps. The receiver receives an input signal representing speech, and produces an output signal representing the input signal after a delay intentionally introduced in the system. The delay corresponds to the length of the buffer. The buffer stores the input sig-nal in the buffer at an input rate, and produces the output signal at an output rate. The buffer controller determines whether the delay or buffer length is greater than a predetermined threshold, and analyzes at least a portion of the input signal stored in the buffer to determine which parts of the input signal may be removed without substantially deterioration of the signal quality accord-ing to at least one predetermined criterion, when the buffer length is greater than a predetermined length. The buffer controller then deletes at least some of the parts of the input signal stored in the buffer that comply with the predetermined criterion.
Description
~ .
AUTOMATIC LENGTH-REDUCING AUDIO DELAY LINE
Technical Field This invention relates generally to communication systems and 5 devices, and more specifically to communication systems using signal buffering and digital signal processing.
Background Several electronic systems may use voice buffers as optional or 10 essential system components to allow time for signal processing, for time-shifting to a more convenient time for transmission or reception, or for performing time-domain processing or sequencing of voice. In each of those systems, the delay caused by voice buffering is both an advantage and a disadvantage. Typically, the delay is an advantage at the beginning 15 of a transmission (e.g., to obtain a communication channel), but a disadvantage at the end of a transmission because of the increased turn-around time between the parties communicating. Thus, a need exists for a method for automatically reducing the intentional delay toward zero (i.e., substantially eliminating the delay).
Summary of the Invention Briefly, according to the invention, an electronic system and a method substantially reduces any delays caused by buffering of a signal containing speech. The electronic system comprises a buffer for storing 25 signals representing speech, and a controller coupled to the buffer for controlling the buffer. The electronic system substantially reduces any delays in the output of the buffer utilizing a method in accordance with the invention. The method comprises the following steps. The buffer receives the input signal representing speech, and produces an output 30 signal representing the input signal, possibly after a delay. The buffer stores the input signal in the buffer at an input rate, and produces an output signal representing the input signal, at an output rate. The buffer .~, ~
controller determines whether the current buffer length is greater than some predetermined minimum threshold, and analyzes at least a portion of the input signal stored in the buffer to determine which parts of the input signal may be removed from the stored signal. A predetermined 5 criterion or set of criteria may be used to determine which parts of the stored signal may be removed without substantial distortion of the stored signal. According to a possible criterion, the buffer controller determines which parts of the stored signal have a magnitude less than a predetermined energy level, when the buffer controller determines that 10 the current buffer length is greater than some predetermined minimum threshold, and those parts are then deleted. According to another aspect of the invention, the criterion for removing parts of the stored signal is whether those parts of the stored signal are repetitive (i.e., representing pitch periods of the stored speech).
Brief Description of the Drawings Figure 1 is a simplified block diagram showing an audio-producing system in accordance with the invention.
Figure 2 is a flow chart illustrating an output process in one 20 embodiment of the invention.
Figure 3 is a flow chart illustrating a process for calculating the energy in a segment of the buffer.
Figures 4A and 4B show a flow chart illustrating an output process in another embodiment of the invention.
Detailed Descri~tion of the Preferred Embodiment Referring ~o Figure 1, a simplified block diagram is shown, depicting a communication device 10 incorporating the invention. An antenna 12 receives radio signals containing speech information. When a 30 transmit/receive switch 14 is in the receive position, the radio signals are provided to a conventional receiver section 16 which, in turn, produces an input signal representing speech information that is provided to an ~a8~47~
analog-to-digital converter 22, for producing a digital signal at an input rate set by a controller 28 (e.g., a microprocessor, a digital signal processor, or custom hardware specifically designed to perform the functions of the invention). The digital signal is then sampled and stored in a conventional circular buffer 26 that comprises a memory (e.g., a large dynamic random access memory DRAM). The controller 28 causes the buffer 26 to sample and store the input signal (at an input rate), and provides an input rate clock signal to the analog-to-digital converter 22, and an output rate clock signal to the digital-to-analog converter 30. The A/D converter 22 includes the proper anti-aliasing filter and the D/A
converter 30 includes the proper reconstruction filter as is well known in the art. The clock signals set the input and output rates (respectively) of the buffer 26. Moreover, the controller 28 controls the memory address to cause successive samples to be stored in an increasing address of memory, and additionally controls a similar process to read samples from the buffer 26 to produce a digital audio signal, and convert it to a restored analog signal, with good fidelity, with the digital-to-analog converter 30, for presentation to a listener by a speaker 34 after processing by a conventional audio section 32. The rate at which the cligital audio signals are read from the buffer 26 and converted to analog form, with the digital-to-analog converter 30, is controlled by the con~roller 28. In addition, the controller 28 can offset the read address from the write address of the audio stored in the buffer 26, to produce a delay that is well known.
The radio 10 also includes a transmitter section which transmits information received at an input 38 (e.g., a microphone), when the switch 14 is in the transmit position. An amplifier 33 is coupled to the microphone 38 to provide an amplified signal to an analog-to-digital converter 31 which converts the received signal to digital form. The digital signal provided by the A/D converter 31 is stored in a buffer 27 at an input rate controlled by the controller 28. Similarly to the case of the ~, receiver, the buffer output is converted to analog for by a D/A converter ~`
~ 1~ 208947~
23. The A/D converter 31 includes the proper anti-aliasing filter and the D/A converter 23 includes the proper reconstruction filter. The resulting analog signal is transmitted by a conventional transmitter 18 when the T/R switch 14 is in the transmit position. The transmission mode of the radio 10 is triggered by activating a push-to-talk (PTT) switch 36, or in response to input from voice recognition unit 37.
A voice buffer (such as buffers 26 and 27) may be implemented with a moving magnetic tape and fixed heads, fixed tape and moving heads, a combination of these, or a random access memory (RAM) with a read address and a write address. The reading and writing processes in the buffer can be controlled by read and write controllers, which could be advanced at the same rate, at different rates or either one or both can be stopped. The read and write controllers could be a part of the controller 28. The distance between the write and read positions, expressed in time, represents the delay in the presentation (i.e., reading) of the stored voice by the buffer from the time that the voice was stored (i.e., written) in the buffer. For example, a tape moving at a rate of 10 in./sec., with heads one inch apart will have a delay of 100 msec. The delay may also be called the "current buffer length". The maximum buffer length of the buffer RAM is the total number of samples it can hold divided by the sample rate (e.g., 1.0 Megabyte/10,000 samples/sec. x 1 sample/byte =
100 sec.). If there is complete control over the present positions of the read and write pointers and over the rate at which they advance, a variable-length buffer can be realized.
According to an embodiment of the invention, when the PTT
switch 36 is pressed, the controller 28 generates a data packet to be sent at the beginning of the transmission. This packet is called the PTT
ID, and may last about 600 msec. During this time, the audio signal at the microphone 38 is recorded by the buffer 27. After the PTT ID has been sent, the audio signal stored in the buffer 27 is played back, now delayed by the hold off time .,~ ~
WO 9~/06467 PCI'tUS91/0721S
~ 5 2~8~78 (e.g., about 600 msec.). Thus, no audio has been lost at the beginning of the transmission. According to the invention, the controller 28 analyzes the buffer 27, and c~ses the delay to be reduced by deleting at least some portions of the stored signal 5 that comply with a set of predetermined criteria. The predetermined criteria include (1 ) whether the parts to be removed correspond to p~uses in the speech represented by the stored audio signal (i.e., parts of the stored signal having low signal energy), or (2) whether ths parts to be removed correspond 10 to repetitive segments of the stored signal (i.e., pitch periods).
Complete reduction of the delay may be desirable but is not necessary. For example, if the described 600 millisecond delay of a 15 second transmission can be reduced to 100 milliseconds, the turn around would be greatly improved.
Thus, it is unnecessary to remove all of the parts of the stored information signal complying with the predetermined criterion (or criteria). In the case that the portions of the stored speech to be removed represent silence gaps in the speech, the duration of those gaps may be shortened by some predetermined 20 percentage, thereby reducing the delay in the buffer with little loss of audio quality. In cases of pitch period removal, the pitch information of the stored voice signal can also be determined by processing the buffer data. Based on this information, the removal of a selected pitch period could be made. Additionally, a 2~ combination of pitch period removal and voice gap removal may be performed to improve performance of the buffer length reduction method.
According to another embodiment of the invention, a voice recognition unit 37 may be included in the radio 10 instead of the 30 PTT switch 36. The voice recognition unit 37 is coupled to the controller 28 via an address and data bus. In this embodiment, the voice recognition unit 37 analyzes the speech samples received at the microphone 38. Since a certain amount of time is required to process the received speech, a voice buffer in which 35 to store received signals is required. If the voice recognition unit 37 determines that the voice signal coming into the microphone is 2~8~478 ~
to be lransn,ill6~, the controller 28 is informed, causing the unit to begin transmitting. The transmit audio is played back at the proper starting point. This starting point will represent audio information already stored in the buffer. thus, no audio is lost at the beginning of the transmission. The controller analyzes the buffer 27 and causes the delay to be reduced by using a speech-processing method. For example, one method would involve looking for silence gaps in the stored speech signal and reducing the the duration of those silence gaps by some percentage, thereby reducing the delay in the buffer with little effect on audio quality. The pitch information of the voice could also be determined by processing the buffer data. Based on this information, the removal of a carefully selected pitch period could be made, thus reducing the buffer length with minimal effect on the voice quality of the system. Since the removal of pitch information implies that there is a minimum buffar length, a combination of pitch removal and silence gap removal could be performed to improve the performance of the buffer reduction method.
Intentional delays in production of a voice buffer output are useful in both the receive and transmit paths of the radio 10. Thus the delay reduction processes of the invention are also useful in both paths. The only difference between those applications is in the source of the high-level decision as to whether the output process should occur.
Referring to Figure 2, there is shown a flow chart of the output process of one method for buffer output delay reduction in accordance with the invention. Step 100 is entered when it is time to produce an output sample. This may be implemented in a microprocessor or a digital signal processor by having a timer interrupt the normal instruction execution. In step 102, an interrupt timer is set up to produce an interrupt at the next output sample time. A decision 104 is then made to determine whether the read address of the buffer equals the write address of the buffer (i.e., whether the buffer is empty). If the read address of the buffer equals the write address of the buffer (i.e., the buffer is WO 92/06467 PCI~/US91/07215 7 ~08~47~
empty) the method proceeds to step 106 to return to the main instruction stream to wait for an input sample to be stored. If the read address of the buffer does not equal the write address of the bu~fer, a further decision 108 must be made as to whether the 5 read address modulo window size equals zero (i.e., whether it is time to examine the buffer contents). If it is zero, the current length of the buffer is c~lcul~ted in step 110, with the controller. If the read address modulo window size does not equal zero, the method proceeds to step 126.
After step 110, a decision 112 on whether the current length of the circular buffer is greater than the window size. If it is greater, the energy in the window is calculated (114). A decision 116 then determines whether the energy in the window is below a predetermined threshold that indicates that a pause in the speech 15 represented by the stored signal may be occurring. A portion of such pauses may be eliminated from the stored signal without significant loss of quality. If the energy in the window is below the predetermined threshold, the window size minus one is added to the read address, in step 118. If the current length of the circular 20 buffer is not greater than the window size, or the energy in the window is not greater than the predetermined threshold, the method proceeds to step 126.
A~ter step 118, a decision 122 is made to deterrnine whether the read address resulting from step 118 is less than the 2~ end address. If it is not less than the end address, the process proceeds to step 126. If the read address resulting from step 118 is less than the end address, the read address is set to be equal to the begin address, and the process then proceeds to step 126.
In step 126 the controller 28 reads the memory at the read 30 address, and writes it to the D/A converter 23 (or 30). The read address is then incremented by one in step 128. A decision 130 is then made to determine whether the incremented read address is greater than the end address. If it is, the read address is reset to be the begin address, in step 132, and the operation in step 3~ 106 will cause the normal processing to continue, until it is time to produce a new output sample. In the event that decision 130 ~8g47~ 8 ~ .
determines that the incremented read address is not greater than the end address, the buffer returns to normal operation in step 106.
Referring to Figure 3, there is shown a flow chart illustrating 5 a process 200 for calculating the energy in a segment of the buffer. In step 202 a sample counter parameter (i~ is set at zero, a sum of squares parameter is set at 0.0, and an address parameter is set at the read address. A decision 206 is then made as to whether the current address is greater than the end address. to 10 determine whether the segment of samples has wrapped around the physical end of memory. If the current address is not greater than the end address, the process proceeds to step 210. If it is greater than the end address, then the address parameter is set at the begin address, in step 208. The contents of the memory 15 address are squared and added to the current value of the sum of squares parameter, in step 210. The contents of the memory at the current address are several voltage levels. Their squares approximate a measure of the energy contents. The sample parameter is then incremented by one in step 212. A decision 20 214 is then made to determine whether the sarnple counter equals the window size. The window size represents the part of the signal that may be deleted. If the sample counter is less than the window size, the process returns to step 206. If the sample counter is not less than the window size, the process continues 25 (216), where the sum of squares will be returned to the calling routine.
Referring to Figures 4A and 4B there is shown a flow chart illustrating an output process in another embodiment of the invention. According to this embodiment the buffer output delay is 30 shortened by deleting at least a portion of the buffered speech signal that is repetitive (i.e., having a well behaved pitch characteristic), with minimal degradation in the audio quality of the output signal. The process begins when an interrupt 300 occurs for the output process of the buffer. In step 302 an output 35 interruption timer is set up. Then in decision 302 a determination is made as to whether the read address equals the write address.
;
WO 92/06467 PCI~/US91/0721!;
9 2~89478 If the read ~dress is not equal to the write address, that signifies that the buffer output has some delay. Therefore in step 308, the length of the buffer is calculated when thers is a delay in the output process . If the read address equals the write address, that 5 signifies that the buffer is empty and the process returns (306) to the main instruction stream at the point where it was interrupted.
In decision 310 a determination is made on whether the buffer length is greater than a predetermined threshold. If it is not, that means that the subject portion of stored speech is not suitable for 10 removal and the process proceeds to step 322. If it is, a further decision 312 is made on whether the read address modulo window size equals zero. This occurs every multiple of the window size. If the modulo function is zero, then in step 314 the pitch period is determined (i.e., the beginning and end of a pitch 15 period to be removed). If the modulo function is not zero, the process proceeds to decision 316. In decision 316 it is determined whether the pitch period is suitable for removal (i.e., a "True" condition). If a True condition exists in step 316, a further decision (318) is made. If a True condition does not exist, then 20 the process goes to step 322. In decision 318 it is determined whether the rea~ address is greater than or equal to the pitch period start and less than the pitch period stop. If it is not, that indicates that the read address is not in the range of the part of the stored signal to be removed, and the process goes to step 25 322. If on the other hand, the read address is greater than or equal to the pitch start and less than the pitch stop, that means that a portion of the stored speech has been identified for removal and the process goes to step 320. In step 320 a read address pointer is set at the pitch period stop address (i.e., the pointer 30 skips over the parts of the stored speech not marked for removal).
In step 322 the memory at the read address is read and written to the D/A converter at the output of the buffer (i.e., the output process continues). The process pointer is then incremented to the next location, in step 324. A decision 326 is made as to 35 whether the current read address is greater than the end address (i.e., whether or not the physical end of memory has been .
~ '?~ 10 ~089478 reached). If it is, the current read address is set to be equal to the begin address ~step 328) and in step 330 the process returns to the main instruction stream at the point where it was interrupted. If the current read address is not greater than the end address, step 328 is skipped.
Steps 326 and 328 form the circular buffer.
Technical Field This invention relates generally to communication systems and 5 devices, and more specifically to communication systems using signal buffering and digital signal processing.
Background Several electronic systems may use voice buffers as optional or 10 essential system components to allow time for signal processing, for time-shifting to a more convenient time for transmission or reception, or for performing time-domain processing or sequencing of voice. In each of those systems, the delay caused by voice buffering is both an advantage and a disadvantage. Typically, the delay is an advantage at the beginning 15 of a transmission (e.g., to obtain a communication channel), but a disadvantage at the end of a transmission because of the increased turn-around time between the parties communicating. Thus, a need exists for a method for automatically reducing the intentional delay toward zero (i.e., substantially eliminating the delay).
Summary of the Invention Briefly, according to the invention, an electronic system and a method substantially reduces any delays caused by buffering of a signal containing speech. The electronic system comprises a buffer for storing 25 signals representing speech, and a controller coupled to the buffer for controlling the buffer. The electronic system substantially reduces any delays in the output of the buffer utilizing a method in accordance with the invention. The method comprises the following steps. The buffer receives the input signal representing speech, and produces an output 30 signal representing the input signal, possibly after a delay. The buffer stores the input signal in the buffer at an input rate, and produces an output signal representing the input signal, at an output rate. The buffer .~, ~
controller determines whether the current buffer length is greater than some predetermined minimum threshold, and analyzes at least a portion of the input signal stored in the buffer to determine which parts of the input signal may be removed from the stored signal. A predetermined 5 criterion or set of criteria may be used to determine which parts of the stored signal may be removed without substantial distortion of the stored signal. According to a possible criterion, the buffer controller determines which parts of the stored signal have a magnitude less than a predetermined energy level, when the buffer controller determines that 10 the current buffer length is greater than some predetermined minimum threshold, and those parts are then deleted. According to another aspect of the invention, the criterion for removing parts of the stored signal is whether those parts of the stored signal are repetitive (i.e., representing pitch periods of the stored speech).
Brief Description of the Drawings Figure 1 is a simplified block diagram showing an audio-producing system in accordance with the invention.
Figure 2 is a flow chart illustrating an output process in one 20 embodiment of the invention.
Figure 3 is a flow chart illustrating a process for calculating the energy in a segment of the buffer.
Figures 4A and 4B show a flow chart illustrating an output process in another embodiment of the invention.
Detailed Descri~tion of the Preferred Embodiment Referring ~o Figure 1, a simplified block diagram is shown, depicting a communication device 10 incorporating the invention. An antenna 12 receives radio signals containing speech information. When a 30 transmit/receive switch 14 is in the receive position, the radio signals are provided to a conventional receiver section 16 which, in turn, produces an input signal representing speech information that is provided to an ~a8~47~
analog-to-digital converter 22, for producing a digital signal at an input rate set by a controller 28 (e.g., a microprocessor, a digital signal processor, or custom hardware specifically designed to perform the functions of the invention). The digital signal is then sampled and stored in a conventional circular buffer 26 that comprises a memory (e.g., a large dynamic random access memory DRAM). The controller 28 causes the buffer 26 to sample and store the input signal (at an input rate), and provides an input rate clock signal to the analog-to-digital converter 22, and an output rate clock signal to the digital-to-analog converter 30. The A/D converter 22 includes the proper anti-aliasing filter and the D/A
converter 30 includes the proper reconstruction filter as is well known in the art. The clock signals set the input and output rates (respectively) of the buffer 26. Moreover, the controller 28 controls the memory address to cause successive samples to be stored in an increasing address of memory, and additionally controls a similar process to read samples from the buffer 26 to produce a digital audio signal, and convert it to a restored analog signal, with good fidelity, with the digital-to-analog converter 30, for presentation to a listener by a speaker 34 after processing by a conventional audio section 32. The rate at which the cligital audio signals are read from the buffer 26 and converted to analog form, with the digital-to-analog converter 30, is controlled by the con~roller 28. In addition, the controller 28 can offset the read address from the write address of the audio stored in the buffer 26, to produce a delay that is well known.
The radio 10 also includes a transmitter section which transmits information received at an input 38 (e.g., a microphone), when the switch 14 is in the transmit position. An amplifier 33 is coupled to the microphone 38 to provide an amplified signal to an analog-to-digital converter 31 which converts the received signal to digital form. The digital signal provided by the A/D converter 31 is stored in a buffer 27 at an input rate controlled by the controller 28. Similarly to the case of the ~, receiver, the buffer output is converted to analog for by a D/A converter ~`
~ 1~ 208947~
23. The A/D converter 31 includes the proper anti-aliasing filter and the D/A converter 23 includes the proper reconstruction filter. The resulting analog signal is transmitted by a conventional transmitter 18 when the T/R switch 14 is in the transmit position. The transmission mode of the radio 10 is triggered by activating a push-to-talk (PTT) switch 36, or in response to input from voice recognition unit 37.
A voice buffer (such as buffers 26 and 27) may be implemented with a moving magnetic tape and fixed heads, fixed tape and moving heads, a combination of these, or a random access memory (RAM) with a read address and a write address. The reading and writing processes in the buffer can be controlled by read and write controllers, which could be advanced at the same rate, at different rates or either one or both can be stopped. The read and write controllers could be a part of the controller 28. The distance between the write and read positions, expressed in time, represents the delay in the presentation (i.e., reading) of the stored voice by the buffer from the time that the voice was stored (i.e., written) in the buffer. For example, a tape moving at a rate of 10 in./sec., with heads one inch apart will have a delay of 100 msec. The delay may also be called the "current buffer length". The maximum buffer length of the buffer RAM is the total number of samples it can hold divided by the sample rate (e.g., 1.0 Megabyte/10,000 samples/sec. x 1 sample/byte =
100 sec.). If there is complete control over the present positions of the read and write pointers and over the rate at which they advance, a variable-length buffer can be realized.
According to an embodiment of the invention, when the PTT
switch 36 is pressed, the controller 28 generates a data packet to be sent at the beginning of the transmission. This packet is called the PTT
ID, and may last about 600 msec. During this time, the audio signal at the microphone 38 is recorded by the buffer 27. After the PTT ID has been sent, the audio signal stored in the buffer 27 is played back, now delayed by the hold off time .,~ ~
WO 9~/06467 PCI'tUS91/0721S
~ 5 2~8~78 (e.g., about 600 msec.). Thus, no audio has been lost at the beginning of the transmission. According to the invention, the controller 28 analyzes the buffer 27, and c~ses the delay to be reduced by deleting at least some portions of the stored signal 5 that comply with a set of predetermined criteria. The predetermined criteria include (1 ) whether the parts to be removed correspond to p~uses in the speech represented by the stored audio signal (i.e., parts of the stored signal having low signal energy), or (2) whether ths parts to be removed correspond 10 to repetitive segments of the stored signal (i.e., pitch periods).
Complete reduction of the delay may be desirable but is not necessary. For example, if the described 600 millisecond delay of a 15 second transmission can be reduced to 100 milliseconds, the turn around would be greatly improved.
Thus, it is unnecessary to remove all of the parts of the stored information signal complying with the predetermined criterion (or criteria). In the case that the portions of the stored speech to be removed represent silence gaps in the speech, the duration of those gaps may be shortened by some predetermined 20 percentage, thereby reducing the delay in the buffer with little loss of audio quality. In cases of pitch period removal, the pitch information of the stored voice signal can also be determined by processing the buffer data. Based on this information, the removal of a selected pitch period could be made. Additionally, a 2~ combination of pitch period removal and voice gap removal may be performed to improve performance of the buffer length reduction method.
According to another embodiment of the invention, a voice recognition unit 37 may be included in the radio 10 instead of the 30 PTT switch 36. The voice recognition unit 37 is coupled to the controller 28 via an address and data bus. In this embodiment, the voice recognition unit 37 analyzes the speech samples received at the microphone 38. Since a certain amount of time is required to process the received speech, a voice buffer in which 35 to store received signals is required. If the voice recognition unit 37 determines that the voice signal coming into the microphone is 2~8~478 ~
to be lransn,ill6~, the controller 28 is informed, causing the unit to begin transmitting. The transmit audio is played back at the proper starting point. This starting point will represent audio information already stored in the buffer. thus, no audio is lost at the beginning of the transmission. The controller analyzes the buffer 27 and causes the delay to be reduced by using a speech-processing method. For example, one method would involve looking for silence gaps in the stored speech signal and reducing the the duration of those silence gaps by some percentage, thereby reducing the delay in the buffer with little effect on audio quality. The pitch information of the voice could also be determined by processing the buffer data. Based on this information, the removal of a carefully selected pitch period could be made, thus reducing the buffer length with minimal effect on the voice quality of the system. Since the removal of pitch information implies that there is a minimum buffar length, a combination of pitch removal and silence gap removal could be performed to improve the performance of the buffer reduction method.
Intentional delays in production of a voice buffer output are useful in both the receive and transmit paths of the radio 10. Thus the delay reduction processes of the invention are also useful in both paths. The only difference between those applications is in the source of the high-level decision as to whether the output process should occur.
Referring to Figure 2, there is shown a flow chart of the output process of one method for buffer output delay reduction in accordance with the invention. Step 100 is entered when it is time to produce an output sample. This may be implemented in a microprocessor or a digital signal processor by having a timer interrupt the normal instruction execution. In step 102, an interrupt timer is set up to produce an interrupt at the next output sample time. A decision 104 is then made to determine whether the read address of the buffer equals the write address of the buffer (i.e., whether the buffer is empty). If the read address of the buffer equals the write address of the buffer (i.e., the buffer is WO 92/06467 PCI~/US91/07215 7 ~08~47~
empty) the method proceeds to step 106 to return to the main instruction stream to wait for an input sample to be stored. If the read address of the buffer does not equal the write address of the bu~fer, a further decision 108 must be made as to whether the 5 read address modulo window size equals zero (i.e., whether it is time to examine the buffer contents). If it is zero, the current length of the buffer is c~lcul~ted in step 110, with the controller. If the read address modulo window size does not equal zero, the method proceeds to step 126.
After step 110, a decision 112 on whether the current length of the circular buffer is greater than the window size. If it is greater, the energy in the window is calculated (114). A decision 116 then determines whether the energy in the window is below a predetermined threshold that indicates that a pause in the speech 15 represented by the stored signal may be occurring. A portion of such pauses may be eliminated from the stored signal without significant loss of quality. If the energy in the window is below the predetermined threshold, the window size minus one is added to the read address, in step 118. If the current length of the circular 20 buffer is not greater than the window size, or the energy in the window is not greater than the predetermined threshold, the method proceeds to step 126.
A~ter step 118, a decision 122 is made to deterrnine whether the read address resulting from step 118 is less than the 2~ end address. If it is not less than the end address, the process proceeds to step 126. If the read address resulting from step 118 is less than the end address, the read address is set to be equal to the begin address, and the process then proceeds to step 126.
In step 126 the controller 28 reads the memory at the read 30 address, and writes it to the D/A converter 23 (or 30). The read address is then incremented by one in step 128. A decision 130 is then made to determine whether the incremented read address is greater than the end address. If it is, the read address is reset to be the begin address, in step 132, and the operation in step 3~ 106 will cause the normal processing to continue, until it is time to produce a new output sample. In the event that decision 130 ~8g47~ 8 ~ .
determines that the incremented read address is not greater than the end address, the buffer returns to normal operation in step 106.
Referring to Figure 3, there is shown a flow chart illustrating 5 a process 200 for calculating the energy in a segment of the buffer. In step 202 a sample counter parameter (i~ is set at zero, a sum of squares parameter is set at 0.0, and an address parameter is set at the read address. A decision 206 is then made as to whether the current address is greater than the end address. to 10 determine whether the segment of samples has wrapped around the physical end of memory. If the current address is not greater than the end address, the process proceeds to step 210. If it is greater than the end address, then the address parameter is set at the begin address, in step 208. The contents of the memory 15 address are squared and added to the current value of the sum of squares parameter, in step 210. The contents of the memory at the current address are several voltage levels. Their squares approximate a measure of the energy contents. The sample parameter is then incremented by one in step 212. A decision 20 214 is then made to determine whether the sarnple counter equals the window size. The window size represents the part of the signal that may be deleted. If the sample counter is less than the window size, the process returns to step 206. If the sample counter is not less than the window size, the process continues 25 (216), where the sum of squares will be returned to the calling routine.
Referring to Figures 4A and 4B there is shown a flow chart illustrating an output process in another embodiment of the invention. According to this embodiment the buffer output delay is 30 shortened by deleting at least a portion of the buffered speech signal that is repetitive (i.e., having a well behaved pitch characteristic), with minimal degradation in the audio quality of the output signal. The process begins when an interrupt 300 occurs for the output process of the buffer. In step 302 an output 35 interruption timer is set up. Then in decision 302 a determination is made as to whether the read address equals the write address.
;
WO 92/06467 PCI~/US91/0721!;
9 2~89478 If the read ~dress is not equal to the write address, that signifies that the buffer output has some delay. Therefore in step 308, the length of the buffer is calculated when thers is a delay in the output process . If the read address equals the write address, that 5 signifies that the buffer is empty and the process returns (306) to the main instruction stream at the point where it was interrupted.
In decision 310 a determination is made on whether the buffer length is greater than a predetermined threshold. If it is not, that means that the subject portion of stored speech is not suitable for 10 removal and the process proceeds to step 322. If it is, a further decision 312 is made on whether the read address modulo window size equals zero. This occurs every multiple of the window size. If the modulo function is zero, then in step 314 the pitch period is determined (i.e., the beginning and end of a pitch 15 period to be removed). If the modulo function is not zero, the process proceeds to decision 316. In decision 316 it is determined whether the pitch period is suitable for removal (i.e., a "True" condition). If a True condition exists in step 316, a further decision (318) is made. If a True condition does not exist, then 20 the process goes to step 322. In decision 318 it is determined whether the rea~ address is greater than or equal to the pitch period start and less than the pitch period stop. If it is not, that indicates that the read address is not in the range of the part of the stored signal to be removed, and the process goes to step 25 322. If on the other hand, the read address is greater than or equal to the pitch start and less than the pitch stop, that means that a portion of the stored speech has been identified for removal and the process goes to step 320. In step 320 a read address pointer is set at the pitch period stop address (i.e., the pointer 30 skips over the parts of the stored speech not marked for removal).
In step 322 the memory at the read address is read and written to the D/A converter at the output of the buffer (i.e., the output process continues). The process pointer is then incremented to the next location, in step 324. A decision 326 is made as to 35 whether the current read address is greater than the end address (i.e., whether or not the physical end of memory has been .
~ '?~ 10 ~089478 reached). If it is, the current read address is set to be equal to the begin address ~step 328) and in step 330 the process returns to the main instruction stream at the point where it was interrupted. If the current read address is not greater than the end address, step 328 is skipped.
Steps 326 and 328 form the circular buffer.
Claims (10)
PROPERTY OF PRIVILEGE IS CLAIMED ARE DEFINED AS FOLLOWS:
1. In an electronic system comprising a buffer having an input for receiving an input signal and an output for providing an output signal representing the input signal after a delay corresponding to a buffer length of the buffer, and a buffer controller coupled to the buffer, a method for substantially reducing any delays in providing the output signal, comprising the steps of:
with the buffer:
(a) storing the input signal in the buffer at an input rate;
(b) producing the output signal at an output rate;
with the buffer controller:
(c) determining whether the current buffer length is greater than a predetermined threshold;
(d) analyzing at least a portion of the input signal stored in the buffer to determine which parts of the input signal comply with at least one predetermined criterion, when the current buffer length is greater than a predetermined threshold; and (e) deleting at least some of the parts of the input signal stored in the buffer that comply with the predetermined criterion, thus decreasing the length of the delay in providing the output signal.
with the buffer:
(a) storing the input signal in the buffer at an input rate;
(b) producing the output signal at an output rate;
with the buffer controller:
(c) determining whether the current buffer length is greater than a predetermined threshold;
(d) analyzing at least a portion of the input signal stored in the buffer to determine which parts of the input signal comply with at least one predetermined criterion, when the current buffer length is greater than a predetermined threshold; and (e) deleting at least some of the parts of the input signal stored in the buffer that comply with the predetermined criterion, thus decreasing the length of the delay in providing the output signal.
2. The method of claim 1, wherein the predetermined criterion comprises whether the magnitude of the energy of the parts to be deleted is a less than the predetermined energy level.
3. The method of claim 1, wherein the predetermined criterion comprises whether the parts to be deleted correspond to a pitch period in the speech represented by the input signal, and the method further comprises the buffer controller steps of:
(c1) comparing a portion of the input signal stored in the buffer to plurality of portions of the input signal previously stored in the buffer and determining the pitch period based on the comparison; and (c2) deleting at least part of the pitch period determined in step (c1), when the current buffer length is greater than a predetermined threshold.
(c1) comparing a portion of the input signal stored in the buffer to plurality of portions of the input signal previously stored in the buffer and determining the pitch period based on the comparison; and (c2) deleting at least part of the pitch period determined in step (c1), when the current buffer length is greater than a predetermined threshold.
4. The method of claim 1, further comprising the step of:
(f) filtering the output signal to remove undesired components produced by deleting the parts of the input signal stored in the buffer that comply with the predetermined criterion.
(f) filtering the output signal to remove undesired components produced by deleting the parts of the input signal stored in the buffer that comply with the predetermined criterion.
5. The method of claim 1, further comprising the step of:
(f) controlling the output rate to be greater than the input rate.
(f) controlling the output rate to be greater than the input rate.
6. The method of claim 5 wherein the output rate is at most ten percent greater than the input rate.
7. The method of claim 3, further comprising the step of:
(f) controlling the output rate to be greater than the input rate, when the portions of the input signal in the buffer representing a pitch period are less than a threshold number.
(f) controlling the output rate to be greater than the input rate, when the portions of the input signal in the buffer representing a pitch period are less than a threshold number.
8. The method of claim 3 wherein step (c1) comprises performing a correlation function.
9. An electronic system comprising:
a buffer for storing an input signal, at an input rate, and for producing an output signal representing the input signal, after a delay corresponding to a buffer length, the output signal being produced at an output rate;
a buffer controller coupled to the buffer, the buffer controller comprising:
means for determining whether the current buffer length is greater than a predetermined threshold;
means for analyzing at least a portion of the input signal stored in the buffer to determine which parts of the input signal have a magnitude less than a predetermined energy level, when the current buffer length is greater than a predetermined threshold; and means for deleting at least some of the parts of the input signal stored in the buffer having a magnitude less than the predetermined energy level, thus deceasing the length of the delay in providing the output signal.
a buffer for storing an input signal, at an input rate, and for producing an output signal representing the input signal, after a delay corresponding to a buffer length, the output signal being produced at an output rate;
a buffer controller coupled to the buffer, the buffer controller comprising:
means for determining whether the current buffer length is greater than a predetermined threshold;
means for analyzing at least a portion of the input signal stored in the buffer to determine which parts of the input signal have a magnitude less than a predetermined energy level, when the current buffer length is greater than a predetermined threshold; and means for deleting at least some of the parts of the input signal stored in the buffer having a magnitude less than the predetermined energy level, thus deceasing the length of the delay in providing the output signal.
10. An electronic device comprising:
a buffer for storing an input signal, at an input rate, and for producing an output signal representing the input signal, after a delay corresponding to a buffer length, the output signal being produced at an output rate;
a buffer controller coupled to the buffer, the buffer controller comprising:
means for determining whether the current buffer length is greater than a predetermined threshold;
comparison means for comparing a portion of the input signal stored in the buffer to plurality of portions of the input signal previously stored in the buffer and determining a pitch period based on the comparison;
analyzer means for analyzing the pitch period determined by the comparison means, when the buffer controller determines that the current buffer length is greater than the predetermined threshold; and means for deleting at least some of the pitch periods of the signal to produce the output signal, thus decreasing the length of the delay in providing the output signal.
a buffer for storing an input signal, at an input rate, and for producing an output signal representing the input signal, after a delay corresponding to a buffer length, the output signal being produced at an output rate;
a buffer controller coupled to the buffer, the buffer controller comprising:
means for determining whether the current buffer length is greater than a predetermined threshold;
comparison means for comparing a portion of the input signal stored in the buffer to plurality of portions of the input signal previously stored in the buffer and determining a pitch period based on the comparison;
analyzer means for analyzing the pitch period determined by the comparison means, when the buffer controller determines that the current buffer length is greater than the predetermined threshold; and means for deleting at least some of the pitch periods of the signal to produce the output signal, thus decreasing the length of the delay in providing the output signal.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US592,001 | 1990-10-01 | ||
| US07/592,001 US5157728A (en) | 1990-10-01 | 1990-10-01 | Automatic length-reducing audio delay line |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CA2089478A1 CA2089478A1 (en) | 1992-04-02 |
| CA2089478C true CA2089478C (en) | 1996-07-02 |
Family
ID=24368860
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA002089478A Expired - Fee Related CA2089478C (en) | 1990-10-01 | 1991-10-01 | Automatic length-reducing audio delay line |
Country Status (6)
| Country | Link |
|---|---|
| US (1) | US5157728A (en) |
| EP (1) | EP0551422B1 (en) |
| JP (1) | JP2850157B2 (en) |
| CA (1) | CA2089478C (en) |
| DE (1) | DE69132029T2 (en) |
| WO (1) | WO1992006467A1 (en) |
Families Citing this family (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH07281689A (en) * | 1994-04-08 | 1995-10-27 | Matsushita Electric Ind Co Ltd | Audio signal transmission device |
| DE69534561T2 (en) * | 1994-08-22 | 2006-07-27 | Sony Corp. | TRANSMITTER-RECEIVER |
| JPH08292798A (en) * | 1995-04-24 | 1996-11-05 | Nec Corp | Voice reproducing control method |
| EP0907161B1 (en) * | 1997-09-18 | 2001-07-11 | Victor Company Of Japan, Ltd. | Apparatus for processing audio signal |
| MY118991A (en) * | 1997-09-22 | 2005-02-28 | Victor Company Of Japan | Apparatus for processing audio signal |
| US6205097B1 (en) | 1999-01-06 | 2001-03-20 | Visteon Global Technologies, Inc. | Method of enhanced data compression rate for a CD player |
| US6243329B1 (en) | 1999-01-06 | 2001-06-05 | Visteon Global Technologies, Inc. | Method of enhanced compression rate for a multi-disc CD player |
| JP4895418B2 (en) * | 1999-08-24 | 2012-03-14 | ソニー株式会社 | Audio reproduction method and audio reproduction apparatus |
| DE10024864A1 (en) * | 2000-05-19 | 2001-11-22 | Siemens Ag | Improving transmission quality in radio communications system after establishing interrupted link involves releasing data separated by time intervals before processing/transmission |
| SE0004839D0 (en) * | 2000-12-22 | 2000-12-22 | Ericsson Telefon Ab L M | Method and communication apparatus in a communication system |
| US6999921B2 (en) * | 2001-12-13 | 2006-02-14 | Motorola, Inc. | Audio overhang reduction by silent frame deletion in wireless calls |
| US7099820B1 (en) * | 2002-02-15 | 2006-08-29 | Cisco Technology, Inc. | Method and apparatus for concealing jitter buffer expansion and contraction |
| US7149412B2 (en) * | 2002-03-01 | 2006-12-12 | Thomson Licensing | Trick mode audio playback |
| US7369671B2 (en) * | 2002-09-16 | 2008-05-06 | Starkey, Laboratories, Inc. | Switching structures for hearing aid |
| US7395208B2 (en) * | 2002-09-27 | 2008-07-01 | Microsoft Corporation | Integrating external voices |
| US20040143675A1 (en) * | 2003-01-16 | 2004-07-22 | Aust Andreas Matthias | Resynchronizing drifted data streams with a minimum of noticeable artifacts |
| US7280813B2 (en) * | 2003-03-13 | 2007-10-09 | Peter Fortier Elliott | Variable delay radio receiver |
| ATE438271T1 (en) * | 2004-03-04 | 2009-08-15 | Ericsson Telefon Ab L M | REDUCING WAITING TIME FOR PRESS TO TALK (PTT) SERVICES |
| US20050227657A1 (en) * | 2004-04-07 | 2005-10-13 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for increasing perceived interactivity in communications systems |
| US20070223660A1 (en) * | 2004-04-09 | 2007-09-27 | Hiroaki Dei | Audio Communication Method And Device |
| US7970020B2 (en) * | 2004-10-27 | 2011-06-28 | Telefonaktiebolaget Lm Ericsson (Publ) | Terminal having plural playback pointers for jitter buffer |
| US20060123194A1 (en) * | 2004-12-02 | 2006-06-08 | Claudio Alex Cukierkopf | Variable effective depth write buffer and methods thereof |
| AU2005334879B2 (en) * | 2005-07-27 | 2009-11-26 | Kabushiki Kaisha Audio-Technica | Conference audio system |
| EP2143103A4 (en) | 2007-03-29 | 2011-11-30 | Ericsson Telefon Ab L M | Method and speech encoder with length adjustment of dtx hangover period |
| GB2451265B (en) * | 2007-07-25 | 2012-01-04 | Wolfson Microelectronics Plc | Control sequencer |
| JP2009288523A (en) * | 2008-05-29 | 2009-12-10 | Toshiba Corp | Speech recognition apparatus and method thereof |
| EP2355497A1 (en) * | 2009-12-21 | 2011-08-10 | Nxp B.V. | Full spectrum time shifting device |
| US9337949B2 (en) | 2011-08-31 | 2016-05-10 | Cablecam, Llc | Control system for an aerially moved payload |
| US10720153B2 (en) * | 2013-12-13 | 2020-07-21 | Harman International Industries, Incorporated | Name-sensitive listening device |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4110560A (en) * | 1977-11-23 | 1978-08-29 | Gte Sylvania Incorporated | Communication apparatus |
| GB2113054A (en) * | 1981-12-21 | 1983-07-27 | Voice Microsystems Ltd | Speech detector |
| US4920534A (en) * | 1986-02-28 | 1990-04-24 | At&T Bell Laboratories | System for controllably eliminating bits from packet information field based on indicator in header and amount of data in packet buffer |
| IL84902A (en) * | 1987-12-21 | 1991-12-15 | D S P Group Israel Ltd | Digital autocorrelation system for detecting speech in noisy audio signal |
-
1990
- 1990-10-01 US US07/592,001 patent/US5157728A/en not_active Expired - Fee Related
-
1991
- 1991-10-01 JP JP3516816A patent/JP2850157B2/en not_active Expired - Lifetime
- 1991-10-01 WO PCT/US1991/007215 patent/WO1992006467A1/en active IP Right Grant
- 1991-10-01 CA CA002089478A patent/CA2089478C/en not_active Expired - Fee Related
- 1991-10-01 DE DE69132029T patent/DE69132029T2/en not_active Expired - Fee Related
- 1991-10-01 EP EP91919362A patent/EP0551422B1/en not_active Expired - Lifetime
Also Published As
| Publication number | Publication date |
|---|---|
| DE69132029T2 (en) | 2000-11-02 |
| CA2089478A1 (en) | 1992-04-02 |
| EP0551422B1 (en) | 2000-03-08 |
| DE69132029D1 (en) | 2000-04-13 |
| EP0551422A4 (en) | 1995-01-25 |
| WO1992006467A1 (en) | 1992-04-16 |
| JP2850157B2 (en) | 1999-01-27 |
| EP0551422A1 (en) | 1993-07-21 |
| US5157728A (en) | 1992-10-20 |
| JPH06501597A (en) | 1994-02-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA2089478C (en) | Automatic length-reducing audio delay line | |
| CA2616305C (en) | Conference audio system | |
| US5675333A (en) | Digital compressed sound recorder | |
| JP4630876B2 (en) | Speech speed conversion method and speech speed converter | |
| JP2538575B2 (en) | Data muting method and apparatus for voice and digital communication systems | |
| US5054078A (en) | Method and apparatus to suspend speech | |
| US7240013B2 (en) | Method and apparatus for controlling buffering of audio stream | |
| EP0817168A1 (en) | Reproducing speed changer | |
| US5293273A (en) | Voice actuated recording device having recovery of initial speech data after pause intervals | |
| JP3378672B2 (en) | Speech speed converter | |
| JP3189587B2 (en) | Audio time base converter | |
| JPH084761Y2 (en) | Receiver for reception of audio and / or video signals | |
| JPH0136119B2 (en) | ||
| JPH08211894A (en) | Voice-grade communication equipment and voice-grade communication system | |
| KR20010085664A (en) | Speech speed converting device | |
| JPH08147874A (en) | Speech speed conversion device | |
| JP2874607B2 (en) | Audio time base converter | |
| JP3607586B2 (en) | Audio delay control device | |
| KR930002585B1 (en) | Voiceless period compression and decompression method | |
| JPS61189600A (en) | Voice storage reproducer | |
| JPH08137500A (en) | Recording and reproducing method for sound or picture signal and device therefor | |
| JPH0832562A (en) | Synchronized transfer circuit for digital signal | |
| JPH0744199A (en) | Speech sound recording and reproducing device | |
| JPS63103457A (en) | Nonaudio suppressing system for voice recorder | |
| JPH06243658A (en) | Reproduction controlling method for recording and reproducing device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| EEER | Examination request | ||
| MKLA | Lapsed |