WO2009087486A2

WO2009087486A2 - Biosensor noise reduction

Info

Publication number: WO2009087486A2
Application number: PCT/IB2008/003960
Authority: WO
Inventors: Emir Delic
Original assignee: Emotiv Systems Pty Ltd
Priority date: 2007-12-31
Filing date: 2008-12-30
Publication date: 2009-07-16
Also published as: WO2009087486A3

Abstract

A method of measuring electroencephalograph (EEG) signals of a user includes generating a first signal from a first electrode in contact with the user's scalp, generating a second signal from a second electrode in proximity to the user's head, the second electrode not contacting the user's skin, and subtracting at least a portion of the second signal from the first signal.

Description

BIOSENSOR NOISE REDUCTION

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Application Serial No. 61/018,182, filed on December 31, 2007.

TECHNICAL FIELD

[0002] This disclosure relates generally to interaction with machines using bio- sensing.

BACKGROUND

[0003] Interactions between humans and machines are usually restricted to the use of input devices such as keyboards, joy sticks, mice, trackballs and the like. Such input devices are cumbersome because they must be manually operated, and in particular operated by hand. Some input devices have been developed to detect eyeball movement or are voice activated to minimize the physical movement required by a user in order to operate these devices. [0004] One area that has been under investigation is the use of bioelectric signals, such as electroencephalograph (EEG) signals, to control machines. Such investigations have generally been limited to clinical and experimental situations. In a clinical application, the electrodes are applied to a patient by a relatively skilled technician. In addition, a patient in a clinical situation is typically stationary, and often in an environment in which introduction and movement of electronic equipment and personnel is limited or controlled. [0005] In general, bio-signal sensing electrodes, particular passive electrodes, are prone to noise and can require noise canceling techniques to achieve satisfactory performance. One noise canceling technique, to minimize impedance at the skin-electrode interface and to minimize interference, involves conditioning the skin where the electrode is to be applied. Typically a scalpel is used to scrape the skin and a liquid disinfectant solution is used to clean the area. Another approach to minimizing impedance and interference at the skin-electrode interface, commonly combined with abrasive and depilatory preparation, is to fill any gap at the interface with a conductive gel or saline solution that can regulate the impedance.

[0006] A conventional apparatus for applying electrodes to a subject's head includes a flexible cap that covers the subject's entire scalp and includes a strap beneath the chin, so that the cap may be snugly secured to the subject's head. This type of apparatus is typically used in a clinical setting and can include over 100 electrodes for some applications.

SUMMARY

[0007] In one aspect, a method of measuring electroencephalograph (EEG) signals of a user includes generating a first signal from a first electrode in contact with the user's scalp, generating a second signal from a second electrode in proximity to the user's head, the second electrode not contacting the user's skin, and subtracting at least a portion of the second signal from the first signal.

[0008] Implementations may include one or more of the following. A plurality of first signals may be generated from a plurality of first electrodes in contact with the users' scalp, the first plurality of signals including the first signal. At least a portion of the second signal may be subtracted from each of the first signals. A plurality of second signals by be generated from a plurality of second electrodes in proximity to the user's head, the plurality of second electrodes not contacting the user's skin, the second plurality of signals including the second signal. Each first signal of the plurality of first signal may be associated with a different second signal, and for each first signal, the associated second signal may be subtracted. There may be an equal number of first signals and second signals. For each particular first signal of the plurality of first signals, a nearest second signal may be subtracted, the nearest second signal being one of the plurality of second signals generated by one of the plurality of second electrodes that is nearest to the particular first electrode that generated the particular first signal. There may be fewer second electrodes than first electrodes. Subtracting the portion of the second signal from the first signal may be performed by analog circuitry or by a programmed processor. A correlation may be determined between the first signal and the second signal. The portion of the second signal subtracted from the first signal may be proportional to the correlation. Whether the correlation exceeds a threshold may be determined. The portion of the second signal subtracted from the first signal may generate a third signal, and a correlation between the third signal and the second signal may be determined, and a portion of the second signal may be subtracted from the third signal. A mental state may be detected using an output form the subtraction of the second signal from the first signal.

[0009] In another aspect, an electroencephalograph (EEG) monitoring system includes a headset to be worn on a user's head, a first electrode supported on the headset to contact the user's scalp when the headset is worn, and a second electrode supported on the headset and configured not to contact the user's skin when the headset is worn, and circuitry configured to subtract the second signal from the first signal.

[0010] Implementations may include one or more of the following. A first plurality of electrodes may generate a first plurality of signals, the first plurality of electrodes including the first electrode. The circuitry may be configured to subtract at least a portion of the second signal from each of the first plurality of signals. A second plurality of electrodes may generate a second plurality of signals, the second plurality of electrodes supported on the headset and configured not to contact the user's skin when the headset is worn, the second plurality of electrodes including the second electrode. The circuitry may be configured to subtract a different one of the second plurality of signals from each of the plurality of first signals. The first plurality of electrodes and the second plurality of electrodes may have an equal number of electrodes. The circuitry may be configured, for each one of the plurality of first signals, to subtract one of the second plurality of signals that is from one of the second plurality of electrodes that is closest to one of the first plurality of electrodes that generated the one of the first plurality of signals. The first plurality of electrodes may have more electrodes than the second plurality of electrodes. The circuitry may be analog circuitry. The analog circuitry may include filters to filter the first and second signals The analog circuitry may include a voltage clamp to cap the voltage of the first and second signals. The analog circuitry may include a differential amplifier to generate an output proportional to the difference between the first signal and the second signal. The circuitry may be a programmed processor. The processor may be programmed to determine a correlation between the first signal and the second signal. The processor may be programmed such that the portion of the second signal subtracted from the first signal is proportional to the correlation. [0011] The processor may be programmed to determine whether the correlation exceeds a threshold. The processor may be on the headset.

[0012] In another aspect, a computer program product, tangibly stored on machine readable medium, has instructions operable to cause a processor to receive a first signal from a first electrode in contact with the user's scalp, receive a second signal from a second electrode in proximity to the user's head, the second electrode not contacting the user's skin, and subtract at least a portion of the second signal from the first signal. [0013] Implementations can realize one or more of the following advantages.

Biosignals, particularly EEG signals, can be captured with reduced noise. This permits an EEG headset to be used more easily and more reliably, particularly in non-clinical applications. [0014] The details of one or more embodiments are set forth in the accompanying drawings and the description below. Other features, objects, and advantages will be apparent from the description and drawings, and from the claims.

DESCRIPTION OF DRAWINGS

[0015] FIG. 1 is a schematic representation of an example signal acquisition system.

[0016] FIG. IA is a schematic representation of a noise compensation system for the signal acquisition system.

[0017] FIG. 2 is a side view of an example headset on a subject's head.

[0018] FIG. 3 A is a schematic representation of an implementation of a noise compensation system.

[0019] FIG. 3B is a schematic representation of an electrode interface.

[0020] FIG. 4A is a schematic representation of another implementation of a noise compensation system.

[0021] FIG. 4B is a flow chart illustrating a method performed by noise compensation software.

[0022] FIG. 5 A is a graph illustrating a scaled reference signal, an EEG signal, and a difference between the EEG signal and the scaled reference signal as a function of time. [0023] FIG. 5B is a graph illustrating a scaled reference signal, an EEG signal, and a difference between the EEG signal and the scaled reference signal as a function of time. [0024] Like reference symbols in the various drawings indicate like elements.

DETAILED DESCRIPTION

[0025] An electrode system to capture bioelectric signals from a subject generally should address various requirements including safety needs, cost, power consumption, performance (e.g., noise), ease of use and subject comfort. However, the relative importance of these factors and the impediments to a satisfactory system may be somewhat different in a non-clinical application than in a clinical application. In particular, in a non-clinical application, the subject is more likely to be moving about, or be in the presence of other moving objects or electronic interference. In addition, in a non-clinical application, the electrodes are more likely to be applied by a person with no training or knowledge of correct application or placement of the electrodes, and thus signal noise becomes more likely. [0026] In some biosensors, the electrodes are subject to capacitively coupled noise.

This capacitively coupled noise depends on sensor contact impedance, e.g., contact impedance of the electrode with the subject's skin, and the location and magnitude of the noise source. Sources of noise can include nearby electronic devices, as well as moving charged bodies in the proximity of the electrode. Objects, people or even body parts that move in the vicinity of the electrode may generate sufficient electrical disturbance to cause noise in the biosensor. In addition, movement of the subject will cause movement of the electrodes relative to the surrounding environment, which can result in noise. [0027] As the contact impedance of a particular sensor increases (e.g., as the contact dries out, or if the contact pulls away from the skin), the susceptibility to noise increases and thus the amplitude of the injected noise becomes larger. In addition, noise across multiple sensors is dependent on the size and direction of the disturbance. So, for example, if hands are clapped to the right of the subject, disturbances in electrodes on the right side of the head will be larger when compared to disturbances in electrodes on the left side. [0028] This capacitively coupled noise can be regarded as common mode noise (with respect to how it is coupled into the system). In general, this noise can be reduced by providing a common mode noise reference sensor that would pick up only common mode noise (e.g., by not making contact with scalp). The common mode noise detected by the sensor can then be subtracted from the EEG channels.

[0029] Turning now to Figure 1 , there is shown a system for detecting and classifying mental states and facial expressions (collectively simply referred to as "states") of a subject and generating signals to represent these states. In general, the system can detect both non- deliberative mental states, for example emotions, e.g., excitement, happiness, fear, sadness, boredom, and other emotions, and deliberative mental states, e.g., a mental command to push, pull or manipulate an object in a real or virtual environment. The system is capable of detection of mental states (both deliberative and non-deliberative) and facial expressions using solely electrical signals, particularly EEG signals, from the subject, and without direct measurement of other physiological processes, such as heart rate, blood pressure, respiration or galvanic skin response, as would be obtained by a heart rate monitor, blood pressure monitor, and the like.

[0030] Signals from the system representing the state of the user are directed to an application 152 running on a computer. The application 152 can also respond to input events by modifying an environment, e.g., a real environment or a virtual environment as displayed on a display 154. Thus, the mental state or facial expressions of a user can be used as a control input for a gaming system, or another application (including a simulator or other interactive environment). [0031] The system includes two main components, a neuro-physio logical signal acquisition device 102, e.g., an electrode headset, that is worn or otherwise carried by a subject, and a state detection engine 114. In brief, the neuro-physio logical signal acquisition device 102 detects bio-signals from the subject, and the state detection engine 114 implements one or more detection algorithms that convert these bio-signals into signals representing the presence (and optionally intensity) of particular states in the subject. [0032] Referring to FIG. IA, the system additionally includes a noise compensation system 40 with one or more sensors 42 to detect external noise, e.g., common mode noise, e.g., noise generated by electronic equipment or moving objects or people in the vicinity of the headset. The noise compensation system 40 subtracts the sensed external noise from the signals from the contact electrodes 103 before the biosignals are analyzed by the state detection engine 114. As discussed below, subtraction of the sensed external noise from the biosignals can be performed by hardware, e.g., analog or digital circuitry, or by software, e.g., as instructions performed by a specialized or general purpose processor. [0033] Returning to FIG. 1, as a headset, the neuro-physiological signal acquisition device 102 includes multiple contact electrodes 103 which, when the headset is properly placed on the subject's head, electrically contact the subject's scalp 10 at predetermined locations to measure EEG signals. It should be noted, however, that the EEG signals measured and used by the system 10 can include signals outside the frequency range, e.g., 0.3-80 Hz, that is customarily recorded for EEG. Moreover, the system 10 can utilize even higher frequencies, e.g., above 100 Hz, for determination of the sensor contact quality. Unlike systems that provide high-resolution 3-D brain scans, e.g., MRI or CAT scans, the headset is generally portable and non-constraining.

[0034] Referring to FIGS. 1 and IA, the neuro-physiological signal acquisition device

102 also includes one or more reference electrodes 42 to detect external noise. Each reference electrode is positioned to pick up only common mode noise and not electrical signals from the subject's skin. Thus, although the reference electrodes 42 are supported on the headset and can be physically close to the contact electrodes 103 (e.g., within several millimeters) and to the user's scalp, none of the reference electrodes contact the subject's scalp. The reference electrodes 42 can be separated from the subject's scalp by an air gap. [0035] In one implementation, there is a single reference electrode for the headset. In this case, the same reference noise signal (albeit with possibly modified amplitudes) can be subtracted from each of the biosignals. In other implementations, there is reference electrode 42 for each contact electrode 103. In this case, each reference electrode 42 can be positioned physically closed to the associated contact electrode 103, and the signal from the reference electrode is subtracted from the signal of the associated contact electrode. In other implementations, there are multiple reference electrodes 42, but fewer reference electrodes than contact electrodes 103. In this case, there may be one reference sensor for each region of the headset (e.g., left side, right side, top, back, front), and the signal from the reference sensor is subtracted from the signal of the contact electrodes in that region. More complicated systems are also possible, in which the noise to subtract from a given contact electrode signal is calculated as a weighted average of signals from multiple reference electrodes.

[0036] Referring to Figure 2, one implementation of an electrode headset 102 is shown. The electrode headset 102 is configured to fit snugly on a subject's head. The headset 102 includes multiple electrode mounts, each configured to mount an electrode. In this implementation the electrode mounts are apertures 130 configured to receive and mount an electrode, e.g., by a press fit (the electrodes and reference electrodes themselves are not illustrated in Figure 2 so as to more clearly illustrate the apertures 130). However, it should be noted that other configurations of electrode mounts can be used. For example, an electrode can be mounted to the electrode headset using a clamp, screw or other suitable connection mechanism and/or configuration.

[0037] In general, the headset 102 can be sufficiently flexible to fit comfortably on the subject's head, but sufficiently rigid to hold the electrodes 103 in approximately the correct positions on the scalp.

[0038] Returning to Figure 1, a headset assembly 100 includes the headset 102 itself, the electrodes 103, and additional circuitry for transmitting the signals from the headset to the state detection engine 114. The signals detected by each of the electrodes 103 on the headset 102 are fed through a sensory interface 104, which can include an amplifier to boost signal strength and a filter to remove noise, and then digitized by an analog-to-digital converters 106. Digitized samples of the signal captured by each of the scalp sensors can be stored in a data buffer 108, such as a memory.

[0039] The data buffer is connected, e.g., through an input/output transmission device

110, such as a wireless 2.4 GHz device, a WiFi or Bluetooth device, to a processing system 120 that runs the state detection engine 114. A microcontroller 109, such as a programmable processor, controls communication between the memory 108 and the input/output device, and handles other aspects of the headset assembly 100, such as controlling power to components to conserve battery lifetime. In particular, if the subtraction of the reference signal is to be performed by software, the instructions can be executed by the microcontroller 109. [0040] The processing system 120 can include a digital signal processor (DSP) 112, to perform desired functional steps of the state detection engine. In general, the DSP 112 performs preprocessing of the digital signals to reduce noise, transforms the signal to "unfold" it from the particular shape of the subject's cortex, and performs the emotion, deliberative mental state and facial expression detection algorithms on the transformed signal. Although illustrated as part of a DSP, the state detection engine can be implemented primarily in hardware using, for example, hardware components such as an Application Specific Integrated Circuit (ASIC), as software, for example, as a memory including a series of instructions to be performed by a DSP or general purpose computer, or using a combination of both software and hardware.

[0041] Systems for detecting mental states are described in U.S. Patent Publication

No. 2007-0173733 and U.S. Patent Publication No. 2007-0066914, both of which are incorporated by reference. Systems for detecting facial expressions are described in U.S. Patent Publication No. 2007-0179396, which is incorporated by reference. [0042] In the illustrated implementation, the head set assembly 100 includes the head set 102, interface 104 and A/D converter(s) 106, MUX/data buffer 108, microcontroller 109, wireless transmission device 110, and signal quality detection circuitry 40, as well as a battery for power supply. In addition, in the illustrated implementation, application 152 and the DSP 112 are part of the same external device 150, e.g., a general purpose computer or a game console.

[0043] However, many other configurations are possible. The state detection engine

114 can be in a dedicated processor unit that is separate from the platform 150 running the application 152. In this case, the processor unit can includes the wireless receiver to receive data from the headset assembly. The processor unit can be connected to the platform 150 by a wired or wireless connection, such as a cable that connects to a USB input of the platform 150. The state detection engine 114 can be software running on the same processor as the application 152. Various components, can be moved onto or off the headset assembly. The buffer 108 could be eliminated or replaced by a multiplexer (MUX), and the data stored directly in the memory of the processing system. A MUX could be placed before the A/D converter stage so that only a single A/D converter is needed. The connection between the head set assembly 100 and the platform 150 can be wired rather than wireless. [0044] As noted above, a noise compensation system 40 senses and subtracts external noise, e.g., common mode noise, from the signals from the contact electrodes 103 before the biosignals are analyzed by the state detection engine 114. [0045] Referring to FIG. 3A, in some implementations, the noise compensation is implemented in hardware. As shown in FIG. 3 A, each contact electrode 103 has an associated reference electrode 42, and subtraction of the common mode is handled in the analog domain at the interface 104. The combination of the EEG electrode 103, reference electrode 42 and components of interface 104 that perform the subtraction can be considered a "smart electrode". In particular, the smart electrode can be a single package that would be mounted in the aperture 130 of the headset 102.

[0046] Referring to FIG. 3B, in some implementations of the smart electrode, the signals from both the EEG electrode 103 and the reference electrode 42 are fed through simple passive RF filters 50a, 50b, to prevent rectification inside the instrumentation amplifier. The amplitude of the input signals is clamped, e.g., by diodes 52a, 52b, to prevent electrostatic discharge damage and amplitude saturation to the amplifier. [0047] A differential amplifier 54, such as an instrumentation amplifier, receives the clamped signal output and amplifies the difference between the EEG electrode 103 and the common mode reference electrode 42. The gain of the amplifier needs to be low enough to ensure that large disturbances do not saturate the output. The output of the smart electrode is then AC or DC coupled to the remainder of the acquisition system. A potential advantage of AC coupling is the potential reduction of wires going to the smart electrode. An instrumentation amplifiers can have the following features: a high common mode rejection ratio, a low input bias current, e.g., in the picoamperage range, a low 1/f noise, and a low current consumption, and stability at small gains.

[0048] Although FIGS. 3 A and 3B illustrate each EEG electrode having an associated reference electrode, as discussed above, there can be fewer reference electrodes than EEG electrodes, in which case the signal from the nearest reference sensor could be routed to the sensor interface 104 of the EEG electrode to be subtracted from the EEG signal. [0049] Referring to FIG. 4A, in some implementations, the noise compensation is implemented in software. As shown in FIG. 4A, the analog signal from each reference electrode can converted to digital form by an A/D converter 106, and the digital signal is directed to a processor, e.g., processor 109. The processor runs noise subtraction software 60 to subtract the common mode signal from the EEG signal.

[0050] A method performed of the noise subtraction software 60 is illustrated in FIG.

4B. In this algorithm, a correlation between the EEG signal and the reference signal is calculated (step 202). This correlation can be calculated using conventional techniques, such as the Pearson product-moment correlation coefficient. [0051] If the absolute value of correlation is determined to be larger than a threshold

(step 204), then the reference signal S_ref_erence is scaled by factor that depends on the correlation to generate a corrected reference signal. For example, the reference signal can simply be multiplied by the correlation, i.e., S_corrected = correlation * S_ref_erence (step 206). The scaled reference signal is then subtracted from the EEG signal (step 208). [0052] Optionally, steps 202-208 can be iterated with the correlation computed again and scaled subtraction performed again until either a maximum number of correlations is reached or the value of correlation is below the threshold (step 210). [0053] Although FIG. 4A illustrates each EEG electrode having an associated reference electrode, as discussed above, there can be fewer reference electrodes than EEG electrodes, in which case the signal from the nearest reference sensor can subtracted by the software from the EEG signal. In addition, although the noise reduction software 60 is illustrated as running on the processor 109 in the headset assembly 100, the noise cancellation could be performed by a processor at another location, such as the processor that runs the signal detection algorithms 114, or in a processor that is separate from but connected to, e.g., by a wired or wireless connection, the processors running the application 152 and/or signal detection algorithms 114.

[0054] FIG. 5 A is a graph illustrating a scaled reference signal, an EEG signal, and a difference between the EEG signal and the scaled reference signal as a function of time in an ideal situation in which each EEG electrode has its own associated reference electrode positioned immediately adjacent the EEG electrode.

[0055] FIG. 5B is a graph illustrating a scaled reference signal, an EEG signal, and a difference between the EEG signal and the scaled reference signal as a function of time in a system in which the headset has only a single reference electrode (and thus the reference electrode positioned cannot be immediately adjacent the EEG electrode). Without being limited to any particular theory, the different positions of noise sources relative to the EEG electrodes can cause a phase delay of the common mode noise signal that is picked by different EEG sensors relative to the common mode noise signal picked up by the reference sensor.

[0056] A potential drawback of a hardware implementation is cost, but the hardware approach is less likely to result in tolerance issues between channels and can provide a higher degree of common mode noise rejection than a software algorithm based solution. [0057] Variations in the analog filter components can introduce distortions in the measured noise and limit the maximum common mode noise attenuation. To compensate for these variation, a calibration sequence can be performed before the system is operated by the subject, e.g., the calibration sequence can be performed by the manufacturer. In the calibration sequence, the response of the filter to a known noise course is measured, and then this information is used to appropriately normalize all the EEG channels. Alternatively, these variations can simply be ignored, which provides reduced accuracy in the EEG signal sent to the detection algorithms, but can have a simpler and easier to implement noise cancellation algorithm.

[0058] Embodiments of the invention and all of the functional operations described in this specification can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the structural means disclosed in this specification and structural equivalents thereof, or in combinations of them. Embodiments of the invention can be implemented as one or more computer program products, i.e., one or more computer programs tangibly embodied in an information carrier, e.g., in a machine readable storage device or in a propagated signal, for execution by, or to control the operation of, data processing apparatus, e.g., a programmable processor, a computer, or multiple processors or computers. A computer program (also known as a program, software, software application, or code) can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a stand alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program does not necessarily correspond to a file. A program can be stored in a portion of a file that holds other programs or data, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub programs, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers at one site or distributed across multiple sites and interconnected by a communication network.

[0059] The processes and logic flows described in this specification can be performed by one or more programmable processors executing one or more computer programs to perform functions by operating on input data and generating output. The processes and logic flows can also be performed by, and apparatus can also be implemented as, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit).

[0060] A number of embodiments of the invention have been described.

Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the invention. For example, an advanced smart electrodes could subtract a scaled amount of the reference signal from the EEG signal through use of feedback techniques. Accordingly, other embodiments are within the scope of the following claims.

Claims

WHAT IS CLAIMED IS:

1. A method of measuring electroencephalograph (EEG) signals of a user, comprising: generating a first signal from a first electrode in contact with the user's scalp; generating a second signal from a second electrode in proximity to the user's head, the second electrode not contacting the user's skin; and subtracting at least a portion of the second signal from the first signal.

2. The method of claim 1 , further comprising generating a plurality of first signals from a plurality of first electrodes in contact with the users' scalp, the first plurality of signals including the first signal.

3. The method of claim 2, further comprising subtracting at least a portion of the second signal from each of the first signals.

4. The method of claim 2, further comprising generating a plurality of second signals from a plurality of second electrodes in proximity to the user's head, the plurality of second electrodes not contacting the user's skin, the second plurality of signals including the second signal.

5. The method of claim 4, wherein each first signal of the plurality of first signal is associated with a different second signal, and the method further comprises, for each first signal, subtracting the associated second signal.

6. The method of claim 5, wherein there are an equal number of first signals and second signals.

7. The method of claim 4, further comprising, for each particular first signal of the plurality of first signals, subtracting a nearest second signal, the nearest second signal being one of the plurality of second signals generated by one of the plurality of second electrodes that is nearest to the particular first electrode that generated the particular first signal.

8. The method of claim 7, wherein there are fewer second electrodes than first electrodes.

9. The method of claim 1 , wherein subtracting the portion of the second signal from the first signal is performed by analog circuitry.

10. The method of claim 1 , wherein subtracting the portion of the second signal from the first signal is performed by a programmed processor.

11. The method of claim 1 , further comprising determining a correlation between the first signal and the second signal.

12. The method of claim 11 , wherein the portion of the second signal subtracted from the first signal is proportional to the correlation.

13. The method of claim 11 , further comprising determining whether the correlation exceeds a threshold.

14. The method of claim 12, wherein subtracting the portion of the second signal from the first signal generates a third signal, and the method further comprises determining a correlation between the third signal and the second signal and subtracting a portion of the second signal from the third signal.

15. The method of claim 1, further comprising detecting a mental state using an output form the subtraction of the second signal from the first signal.

16. An electroencephalograph (EEG) monitoring system, comprising: a headset to be worn on a user's head; a first electrode supported on the headset to contact the user's scalp when the headset is worn; a second electrode supported on the headset and configured not to contact the user's skin when the headset is worn; and circuitry configured to subtract the second signal from the first signal.

17. The system of claim 16, further comprising a first plurality of electrodes to generate a first plurality of signals, the first plurality of electrodes including the first electrode.

18. The system of claim 17, wherein the circuitry is configured to subtract at least a portion of the second signal from each of the first plurality of signals.

19. The system of claim 17, further comprising a second plurality of electrodes to generate a second plurality of signals, the second plurality of electrodes supported on the headset and configured not to contact the user's skin when the headset is worn, the second plurality of electrodes including the second electrode.

20. The system of claim 19, wherein the circuitry is configured to subtract a different one of the second plurality of signals from each of the plurality of first signals.

21. The system of claim 20, wherein the first plurality of electrodes and the second plurality of electrodes have an equal number of electrodes.

22. The system of claim 19, wherein the circuitry is configured, for each one of the plurality of first signals, to subtract one of the second plurality of signals that is from one of the second plurality of electrodes that is closest to one of the first plurality of electrodes that generated the one of the first plurality of signals.

23. The system of claim 22, wherein the first plurality of electrodes has more electrodes than the second plurality of electrodes.

24. The system of claim 16, wherein the circuitry comprises analog circuitry.

25. The system of claim 24,wherein the analog circuitry includes filters to filter the first and second signals

26. The system of claim 25, wherein the analog circuitry includes a voltage clamp to cap the voltage of the first and second signals.

27. The system of claim 24, wherein the analog circuitry includes a differential amplifier to generate an output proportional to the difference between the first signal and the second signal.

28. The system of claim 16, wherein the circuitry comprises a programmed processor.

29. The system of claim 28, wherein the processor is programmed to determine a correlation between the first signal and the second signal.

30. The system of claim 29, wherein the processor is programmed such that the portion of the second signal subtracted from the first signal is proportional to the correlation.

31. The system of claim 29, wherein the processor is programmed to determine whether the correlation exceeds a threshold.

32. The system of claim 28, wherein the processor is on the headset.

33. A computer program product, tangibly stored on machine readable medium, the product comprising instructions operable to cause a processor to: receive a first signal from a first electrode in contact with the user's scalp; receive a second signal from a second electrode in proximity to the user's head, the second electrode not contacting the user's skin; and subtract at least a portion of the second signal from the first signal.