JP7660416B2 - Electroencephalogram signal processing method, device, program, and vehicle seat - Google Patents

Description

The present disclosure relates to an electroencephalogram signal processing method, an electroencephalogram signal processing device, an electroencephalogram signal processing program, and a vehicle seat.

In the technology described in Patent Document 1, an electroencephalogram (EEG) signal emitted by the driver while the vehicle is traveling is acquired by an electroencephalogram (EEG) measuring device. In addition, the occurrence of biological artifacts (artifacts caused by body movement) caused by, for example, the driver's head shaking is detected, for example, by a camera or by detecting abnormally large EEG voltage. In addition, a filter corresponding to the detected biological artifact is selected, for example, from the memory of a controller. Then, using the selected filter, a true EEG signal is acquired from the EEG signal acquired by the EEG measuring device, with the influence of the biological artifacts removed.

JP 2017-42261 A

However, the technology described in Patent Document 1 has the disadvantage that it requires a camera or the like to detect body movements, making the configuration complicated. In addition, the technology described in Patent Document 1 has the disadvantage that, among the various artifacts contained in the EEG signal, it can only reduce predetermined biological artifacts (artifacts for which a corresponding filter has been created in advance and stored in memory).

For example, the technology described in Patent Document 1 is unable to reduce artifacts caused by body movements that are not anticipated, such as simultaneous blinking, yawning, and head shaking, because it does not store a filter that corresponds to the artifacts caused by the body movements. Also, for example, the technology described in Patent Document 1 does not take into account artifacts caused by the environment or artifacts that occur during communication that are included in the EEG signal, and is therefore unable to reduce artifacts caused by the environment or artifacts that occur during communication.

The present disclosure has been made in consideration of the above facts, and aims to provide an EEG signal processing method, an EEG signal processing device, an EEG signal processing program, and a vehicle seat that can reduce various artifacts contained in EEG signals without complicating the configuration.

The EEG signal processing method according to the first aspect acquires an EEG signal from a subject, performs an abnormal value reduction process on the acquired EEG signal to reduce abnormal values whose amplitude is equal to or greater than a threshold, performs a first filter process on the EEG signal that has undergone the abnormal value reduction process by applying a smoothing filter, performs a DC offset reduction process on the EEG signal that has undergone the first filter process by reducing a DC offset component, and performs a second filter process on the EEG signal that has undergone the DC offset reduction process by applying a high-pass filter and a low-pass filter.

In the first aspect, the electroencephalogram signal of the subject is subjected to anomaly reduction processing, first filter processing, DC offset reduction processing, and second filter processing in that order. As a result, as is clear from the results of experiments conducted by the inventors of the present application (described later), it is possible to reduce various artifacts contained in the electroencephalogram signal. In addition, the first aspect does not require a camera or the like for detecting the subject's body movements, and therefore does not lead to a complicated configuration.

In the second aspect, the abnormal value reduction process in the first aspect involves replacing data whose amplitude is equal to or greater than a threshold with the intermediate value of the data before and after it.

According to the second aspect, it is possible to realize an abnormal value reduction process that reduces abnormal values whose amplitude is equal to or exceeds a threshold value through simple processing.

The third aspect is the first or second aspect, in which a significant frequency band in the power spectrum of the EEG signal that has been subjected to the second filter processing is determined, and the state of consciousness of the subject is determined from the determined significant frequency band.

According to the present disclosure, various artifacts contained in an EEG signal can be reduced, and therefore, according to the third aspect, significant frequency bands can be accurately determined from an EEG signal in which various artifacts have been reduced, and the subject's state of consciousness can be accurately determined.

The fourth aspect is to obtain an EEG signal from a subject in a non-resting state in any of the first to third aspects.

According to the fourth aspect, it is possible to obtain an EEG signal from a subject with various artifacts reduced, even when the subject is not in a resting state.

The fifth aspect is any one of the first to fourth aspects, in which an electroencephalogram signal is acquired from a subject who is driving a vehicle.

According to the fifth aspect, it is possible to obtain an EEG signal from a subject with various artifacts reduced, even when the subject is driving a vehicle.

In a sixth aspect, in any one of the first to fifth aspects, the threshold value in the abnormal value reduction process is 150 μV or a value greater than this.

When the subject is not at rest, the magnitude of the artifact caused by body movement is 150 μV or more. In the sixth aspect, by setting the threshold value in the abnormal value reduction process to 150 μV or more, it is possible to appropriately reduce artifacts caused by body movement when the subject is not at rest.

The seventh aspect is any one of the first to sixth aspects, in which first information related to artifacts caused by the subject's body movements contained in the subject's EEG signal is acquired, and at least processing parameters of the abnormal value reduction process are set based on the acquired first information.

The magnitude of artifacts caused by the subject's body movements contained in the subject's EEG signal varies depending on the subject's individual constitution (e.g., how easily they sweat). In the seventh aspect, first information related to artifacts caused by the subject's body movements (e.g., information about the subject's constitution) is acquired, and at least the processing parameters of the abnormal value reduction process are set based on the acquired first information, so that artifacts caused by body movements can be appropriately reduced.

The eighth aspect is any one of the first to seventh aspects, in which second information related to an artifact caused by the environment contained in the electroencephalogram signal of the subject is acquired, and processing parameters of the second filtering process are set based on the acquired second information.

Environmental artifacts contained in the subject's EEG signal vary depending on the environment, such as the location of electromagnetic wave sources in the vicinity. In the eighth aspect, second information (e.g., the ID of the surrounding environment) related to the environmental artifacts is acquired, and processing parameters for the second filter processing are set based on the acquired second information, so that environmental artifacts can be appropriately reduced.

The electroencephalogram signal processing device according to the ninth aspect includes an acquisition unit that acquires an electroencephalogram signal of a subject, an abnormal value reduction unit that performs an abnormal value reduction process on the electroencephalogram signal acquired by the acquisition unit to reduce abnormal values whose amplitude is equal to or greater than a threshold, a first filter unit that performs a first filter process that applies a smoothing filter to the electroencephalogram signal that has undergone the abnormal value reduction process by the abnormal value reduction unit, a DC offset reduction unit that performs a DC offset reduction process that reduces a DC offset component on the electroencephalogram signal that has undergone the first filter process by the first filter unit, and a second filter unit that performs a second filter process that applies a high-pass filter and a low-pass filter to the electroencephalogram signal that has undergone the DC offset reduction process by the DC offset reduction unit.

According to the ninth aspect, similar to the first aspect, various artifacts contained in the EEG signal can be reduced without complicating the configuration.

The vehicle seat according to the tenth aspect includes a seat body on which a driver of the vehicle sits, an EEG measuring unit that measures the EEG of the driver of the vehicle seated in the seat body and outputs an EEG signal, and an EEG signal processing device according to the ninth aspect that acquires the EEG signal output from the EEG measuring unit as the EEG signal of the subject.

According to the tenth aspect, as with the first aspect, it is possible to obtain an EEG signal with reduced various artifacts without complicating the configuration.

The EEG signal processing program according to the eleventh aspect causes a computer to execute processing including acquiring an EEG signal from a subject, performing an abnormal value reduction process on the acquired EEG signal to reduce abnormal values whose amplitude is equal to or greater than a threshold, performing a first filter process on the EEG signal that has undergone the abnormal value reduction process by applying a smoothing filter, performing a DC offset reduction process on the EEG signal that has undergone the first filter process by reducing a DC offset component, and performing a second filter process on the EEG signal that has undergone the DC offset reduction process by applying a high-pass filter and a low-pass filter.

According to the eleventh aspect, similar to the first aspect, various artifacts contained in the EEG signal can be reduced without complicating the configuration.

The present disclosure has the effect of reducing various artifacts contained in EEG signals without complicating the configuration.

1 is a side view of a vehicle seat according to an embodiment; 2 is a block diagram showing a schematic configuration of an electroencephalogram signal processing ECU according to the first embodiment; FIG. FIG. 2 is a functional block diagram of the electroencephalogram signal processing ECU according to the first embodiment. 4 is a flowchart showing electroencephalogram signal processing according to the first embodiment. FIG. 11 is a block diagram showing a schematic configuration of an electroencephalogram signal processing ECU according to a second embodiment. 11 is a table showing an example of a processing parameter DB. 11 is a table illustrating an example of a threshold DB. FIG. 11 is a functional block diagram of an electroencephalogram signal processing ECU according to a second embodiment. 10 is a flowchart showing electroencephalogram signal processing according to the second embodiment. FIG. 1 is a diagram showing a 15 Hz sine wave used as a signal simulating a true electroencephalogram signal in an experiment conducted by the inventors of the present application. FIG. 11 is a diagram showing the power spectrum of the sine wave shown in FIG. 10 . FIG. 2 is a diagram showing a pseudo-electroencephalogram signal created by the present inventors. FIG. 13 is a diagram showing the power spectrum of the pseudo-electroencephalogram signal shown in FIG. 12 . 13 is a diagram showing the power spectrum of an electroencephalogram signal when only an outlier reduction process is performed on a pseudo-electroencephalogram signal. FIG. 13 is a diagram showing the power spectrum of the electroencephalogram signal when only the first filter processing (moving average filter) is performed on the pseudo-electroencephalogram signal. FIG. 13 is a diagram showing the power spectrum of an electroencephalogram signal when only DC offset reduction processing is performed on a pseudo-electroencephalogram signal. FIG. 13 is a diagram showing the power spectrum of the electroencephalogram signal when only the second filter processing (high-pass filter and low-pass filter) is performed on the pseudo-electroencephalogram signal. FIG. 13 is a diagram showing an electroencephalogram signal when an abnormal value reduction process is performed on a pseudo-electroencephalogram signal. FIG. FIG. 19 is a diagram showing the power spectrum of the electroencephalogram signal shown in FIG. 18. 19 is a diagram showing an electroencephalogram signal obtained by performing a first filter process (moving average filter) on the electroencephalogram signal shown in FIG. 18 . FIG. 21 is a diagram showing the power spectrum of the electroencephalogram signal shown in FIG. 20. 21 is a diagram showing an electroencephalogram signal when a DC offset reduction process is performed on the electroencephalogram signal shown in FIG. 20. FIG. FIG. 23 is a diagram showing the power spectrum of the electroencephalogram signal shown in FIG. 22. 23 is a diagram showing the electroencephalogram signal when second filter processing (high-pass filter and low-pass filter) is performed on the electroencephalogram signal shown in FIG. 22. FIG. 25 is a diagram showing the power spectrum of the electroencephalogram signal shown in FIG. 24. This is a line graph showing the power spectrum of an EEG signal when the pseudo-EEG signal is subjected to the following processing in the order of (3) DC offset reduction processing → (1) outlier reduction processing → (2) first filter processing → (4) second filter processing. This is a line graph showing the power spectrum of an EEG signal when the pseudo-EEG signal is subjected to the following processing in the order of (4) second filter processing → (1) outlier reduction processing → (2) first filter processing → (3) DC offset reduction processing. This is a line graph showing the power spectrum of an EEG signal when the pseudo-EEG signal is subjected to the following processing in the order of (2) first filter processing → (1) outlier reduction processing → (3) DC offset reduction processing → (4) second filter processing. This is a line graph showing the power spectrum of an EEG signal when the pseudo-EEG signal is subjected to the following processes in order: (1) outlier reduction processing → (2) first filter processing → (4) second filter processing → (3) DC offset reduction processing. This is a line graph showing the power spectrum of an EEG signal when the pseudo-EEG signal is subjected to the following processing in the order of (2) first filter processing → (3) DC offset reduction processing → (4) second filter processing → (1) abnormal value reduction processing. This is a line graph showing the power spectrum of an EEG signal when the pseudo-EEG signal is subjected to the following processing in the order of (1) outlier reduction processing → (3) DC offset reduction processing → (4) second filter processing → (2) first filter processing. This is a line graph showing the power spectrum of an EEG signal when the pseudo-EEG signal is subjected to the following processing in the order of (1) abnormal value reduction processing → (3) DC offset reduction processing → (2) first filter processing → (4) second filter processing.

An example of an embodiment of the present disclosure is described in detail below with reference to the drawings.

First Embodiment
A vehicle seat 10 shown in Fig. 1 includes a seat body 12. The seat body 12 is provided in a driver's seat of a vehicle, and is seated by a vehicle driver 20, who is an example of a test subject. The driver 20 seated in the seat body 12 operates a steering wheel 22 and pedals (not shown) to drive the vehicle when the vehicle is traveling. In the first embodiment, the vehicle seat 10 is provided in a vehicle of a specific model.

The seat body 12 includes a seat cushion portion 14, a seat back portion 16, and a headrest portion 18. The headrest portion 18 is attached to the upper end portion of the seat back portion 16 in the vertical direction of the vehicle so as to be slidable along the length of the seat back portion 16. The lower end portion of the seat back portion 16 in the vertical direction of the vehicle is attached to the rear end portion of the seat cushion portion 14 in the longitudinal direction of the vehicle via a rotation mechanism (not shown), and is rotatable relative to the seat cushion portion 14 around an axis along the width direction of the vehicle.

The vehicle seat 10 also includes an EEG measuring device 24 and an EEG signal processing ECU (Electronic Control Unit) 26. The EEG measuring device 24 is helmet-shaped and is worn on the head of the driver 20 of the vehicle who is seated in the seat body 12. The EEG measuring device 24 measures the brain waves of the driver 20 while worn on the head of the driver 20, and outputs an EEG signal representing the measured brain waves to the EEG signal processing ECU 26 via wireless communication. The EEG measuring device 24 is an example of an EEG measuring section.

The EEG signal processing ECU 26 is housed in the seat cushion section 14. As shown in FIG. 2, the EEG signal processing ECU 26 includes a CPU (Central Processing Unit) 28, a memory 30 such as a ROM (Read Only Memory) or a RAM (Random Access Memory), and a non-volatile storage section 32 such as a HDD (Hard Disk Drive) or an SSD (Solid State Drive). The EEG signal processing ECU 26 also includes a wireless communication section 34 for wirelessly communicating with the EEG measuring device 24, and an alarm section 36 with a voice output function and a vibration generation function. The CPU 28, the memory 30, the storage section 32, the wireless communication section 34, and the alarm section 36 are connected to each other so as to be able to communicate with each other via an internal bus 38.

The storage unit 32 stores an EEG signal processing program 40. The EEG signal processing ECU 26 functions as each of the functional units shown in FIG. 3 when the EEG signal processing program 40 is read from the storage unit 32 and loaded into the memory 30, and the EEG signal processing program 40 loaded into the memory 30 is executed by the CPU 28. That is, the EEG signal processing ECU 26 functions as an acquisition unit 60, an abnormal value reduction unit 62, a first filter unit 64, a DC offset reduction unit 66, a second filter unit 68, a significant band determination unit 70, and a consciousness state determination unit 72. The EEG signal processing ECU 26 is an example of an EEG signal processing device.

The acquisition unit 60 acquires the EEG signal of the vehicle driver 20 from the EEG measurement device 24. The abnormal value reduction unit 62 performs an abnormal value reduction process on the EEG signal acquired by the acquisition unit 60 to reduce abnormal values whose amplitude is equal to or greater than a threshold. The first filter unit 64 performs a first filter process that applies a moving average filter to the EEG signal that has undergone the abnormal value reduction process by the abnormal value reduction unit 62. Note that the moving average filter is an example of a smoothing filter. The DC offset reduction unit 66 performs a DC offset reduction process on the EEG signal that has undergone the first filter process by the first filter unit 64 to reduce the DC offset component.

The second filter unit 68 performs a second filtering process by applying a high-pass filter and a low-pass filter to the EEG signal that has undergone the DC offset reduction process by the DC offset reduction unit 66. The significant band determination unit 70 determines the significant frequency band in the power spectrum of the EEG signal that has undergone the second filtering process by the second filter unit 68. The consciousness state determination unit 72 determines the consciousness state of the vehicle driver 20 from the significant frequency band determined by the significant band determination unit 70.

Next, as an operation of the first embodiment, the EEG signal processing executed by the EEG signal processing ECU 26 according to the first embodiment while the driver 20 is operating the vehicle will be described with reference to FIG. 4.

In step 100, the acquisition unit 60 acquires an EEG signal representing the brain waves of the driver 20 for a predetermined period of time from the EEG measurement device 24 via wireless communication. When the driver 20 makes a body movement such as blinking or shaking his head, an artifact caused by the body movement is added to this EEG signal. In addition, electromagnetic wave sources (e.g., harnesses, car navigation devices, etc.) are present around the vehicle seat 10 installed in the driver's seat of the vehicle, and the above EEG signal is added with artifacts caused by the environment depending on the arrangement of the surrounding electromagnetic wave sources. In addition, when wireless communication is performed between the EEG measurement device 24 and the EEG signal processing ECU 26, communication artifacts are also added to the above EEG signal.

In step 102, the abnormal value reduction unit 62 determines whether the EEG signal acquired by the acquisition unit 60 includes a portion whose amplitude is equal to or greater than the threshold voltage. In this step 102, it is determined whether the EEG signal includes a portion that includes artifacts due to body movement. In this embodiment, the EEG is acquired from a driver 20 who is not at rest, more specifically, who is operating a vehicle, so that the magnitude of the artifacts due to body movement added to the EEG signal is larger than when the subject is at rest. For this reason, in the first embodiment, the threshold voltage is preset to 150 μV or a value greater than this (200 μV as an example).

If the determination in step 102 is positive, the process proceeds to step 104. In step 104, the abnormal value reduction unit 62 removes data from the EEG signal acquired by the acquisition unit 60 that has an amplitude equal to or greater than the threshold voltage. Then, in step 106, the abnormal value reduction unit 62 interpolates the data removed in step 104 with the median of the data existing before and after the removed data (an intermediate value other than the median may be used), and the process proceeds to step 108. Since artifacts due to body movement have a large amplitude, the artifacts due to body movement are reduced first in steps 102 to 106 (an example of the abnormal value reduction process by the abnormal value reduction unit 62) as described above. Note that if the determination in step 102 is negative, steps 104 and 106 are skipped and the process proceeds to step 108.

In step 108, the first filter unit 64 performs a first filter process in which a moving average filter is applied to the EEG signal that has been subjected to the abnormal value reduction process by the abnormal value reduction unit 62. In the first filter process, high frequency components are suppressed by smoothing the changes in the EEG signal. Note that the section of the moving average filter is preset according to artifacts caused by body movements that are added to the EEG signal in a specific vehicle model in which the vehicle seat 10 is installed.

In step 110, the DC offset reduction unit 66 performs a DC offset reduction process to reduce the DC offset component on the EEG signal that has been subjected to the first filtering process by the first filter unit 64. In the DC offset reduction process, low-frequency components are mainly suppressed in order to reduce artifacts during communication.

In step 112, the second filter unit 68 performs a second filtering process in which the EEG signal that has been subjected to the DC offset reduction process by the DC offset reduction unit 66 is filtered through a high-pass filter and a low-pass filter. Note that the cutoff frequency of the high-pass filter and the cutoff frequency of the low-pass filter are preset according to the frequency band of environmental artifacts that are added to the EEG signal in the specific vehicle model in which the vehicle seat 10 is installed.

The filter waveform in the second filtering process is mountain-shaped, and even if the second filtering process is performed, artifacts that cannot be completely removed remain near the cutoff frequency of the filter. In this embodiment, in order to minimize the artifacts that cannot be completely removed, the first filtering process by the first filter unit 64 and the DC offset reduction process by the DC offset reduction unit 66 are performed in advance.

In this manner, in this embodiment, the EEG signal of the driver 20 is subjected to the following processes in order: abnormal value reduction process by the abnormal value reduction unit 62, first filtering process by the first filter unit 64, DC offset reduction process by the DC offset reduction unit 66, and second filtering process by the second filter unit 68. This makes it possible to obtain an EEG signal in which various artifacts have been reduced, as shown in FIG. 25 as an example.

In step 114, the significant band determination unit 70 performs FFT (Fast Fourier Transform) on the EEG signal from which various artifacts have been reduced to obtain a power spectrum, and performs significant band determination processing to determine significant frequency bands in the obtained power spectrum. This significant band determination processing can be performed using a method in which the frequency band in which the maximum peak occurs on the power spectrum of the EEG signal is determined to be significant, or a method in which the power spectrum of the EEG signal is integrated for each frequency band, such as alpha waves, beta waves, delta waves, and theta waves, and the frequency band in which the integration result shows a maximum value is determined to be significant. Either method may be applied in step 114.

In step 116, the consciousness state determination unit 72 performs consciousness state determination processing to determine the consciousness state of the vehicle driver 20 from the significant frequency band determined by the significant band determination unit 70, and to warn the driver 20 as necessary. Specifically, for example, if the significant frequency band determined by the significant band determination unit 70 is a frequency band corresponding to θ waves, the consciousness state determination unit 72 determines that the driver 20 is in a drowsy consciousness state, and causes the notification unit 36 to output a warning sound or a warning message and generate a vibration. This can wake up the driver 20.

In step 118, the EEG signal processing ECU 26 determines whether the driver 20 has finished driving. If the determination in step 118 is negative, the process returns to step 100, and steps 100 to 118 are repeated until the determination in step 118 is positive. When the driver 20 has finished driving, the determination in step 118 is positive, and EEG signal processing is terminated.

As described above, in the first embodiment, an EEG signal from a subject is acquired, and an abnormal value reduction process is performed on the acquired EEG signal to reduce abnormal values whose amplitude is equal to or greater than a threshold. A first filter process is also performed to apply a smoothing filter to the EEG signal that has undergone the abnormal value reduction process. A DC offset reduction process is also performed to reduce DC offset components on the EEG signal that has undergone the first filter process. Then, a second filter process is performed to apply a high-pass filter and a low-pass filter to the EEG signal that has undergone the DC offset reduction process. This makes it possible to reduce various artifacts contained in the EEG signal without complicating the configuration.

In addition, in the first embodiment, the abnormal value reduction process involves replacing the value of data whose amplitude is equal to or greater than a threshold with the intermediate value of the data before and after it. This makes it possible to realize the abnormal value reduction process, which reduces abnormal values whose amplitude is equal to or greater than a threshold, with simple processing.

In addition, in the first embodiment, a significant frequency band is determined in the power spectrum of the EEG signal that has been subjected to the second filter process, and the subject's state of consciousness is determined from the determined significant frequency band. This makes it possible to accurately determine the significant frequency band from the EEG signal in which various artifacts have been reduced, and therefore to accurately determine the subject's state of consciousness.

In addition, in the first embodiment, an EEG signal is acquired from a subject who is not in a resting state. This makes it possible to obtain an EEG signal from a subject in which various artifacts are reduced, even if the subject is not in a resting state.

In addition, in the first embodiment, an EEG signal is acquired from a subject (driver 20) who is driving a vehicle. This makes it possible to obtain an EEG signal from the subject with various artifacts reduced even when the subject is driving a vehicle.

In addition, in the first embodiment, the threshold value in the abnormal value reduction process is set to 150 μV or a value exceeding this. This makes it possible to appropriately reduce artifacts due to body movement when the subject is not in a resting state.

Second Embodiment
Next, a second embodiment of the present disclosure will be described. Note that the same parts as those in the first embodiment are denoted by the same reference numerals, and the description thereof will be omitted.

In the second embodiment, the vehicle seat 10 is installed in a vehicle of any one of a plurality of vehicle types. As shown in FIG. 5, the brainwave signal processing ECU 26 according to the second embodiment is connected via a communication line to a vehicle-side ECU 50 mounted in the vehicle in which the vehicle seat 10 is installed (hereinafter referred to as the "specific vehicle"). The vehicle-side ECU 50 includes a CPU, memory, storage unit, etc., similar to the brainwave signal processing ECU 26, and the storage unit of the vehicle-side ECU 50 stores a vehicle model ID representing the vehicle model of the specific vehicle. The brainwave signal processing ECU 26 acquires the vehicle model ID of the specific vehicle from the vehicle-side ECU 50.

The memory unit 32 of the EEG signal processing ECU 26 stores a processing parameter DB 42. In the second embodiment, environmental artifacts added to the EEG signal are measured in advance for each vehicle model, and processing parameters are determined in advance for each vehicle model according to the artifact measurement results. Here, the processing parameters include the cutoff frequency of the high-pass filter and the cutoff frequency of the low-pass filter in the second filter process by the second filter unit 68. As shown in FIG. 6, the processing parameters determined for each vehicle model are registered in the processing parameter DB 42 in association with the vehicle model ID.

In the second embodiment, the wireless communication unit 34 is capable of wireless communication with a wearable device 52, such as a smart watch, worn by the driver 20. The wearable device 52 has a function of detecting the physical constitution of the driver 20, such as the level of sweatiness, and transmits the physical constitution of the driver 20, including the level of sweatiness, detected by this function, to the EEG signal processing ECU 26 as physical constitution information of the driver 20.

The memory unit 32 of the EEG signal processing ECU 26 stores a threshold DB 44. In the second embodiment, a threshold for the abnormal value reduction process by the abnormal value reduction unit 62 is determined in advance for each physical constitution of the driver 20 (e.g., the level of sweatiness, etc.). As shown in FIG. 7, the threshold DB 44 stores the threshold determined for each physical constitution of the driver 20 (e.g., the level of sweatiness, etc.) in association with physical constitution information of the driver 20.

In the second embodiment, the EEG signal processing ECU 26 includes a communication unit 46 (see FIG. 5) for communicating with the vehicle-side ECU 50. The EEG signal processing ECU 26 also functions as a parameter setting unit 74 shown in FIG. 8. The parameter setting unit 74 performs steps 90 to 96 of the EEG signal processing according to the second embodiment (see FIG. 9).

That is, in step 90, the parameter setting unit 74 acquires the vehicle model ID of the specific vehicle from the vehicle-side ECU 50. In step 92, the parameter setting unit 74 reads out the processing parameters registered in the processing parameter DB 42 in association with the vehicle model ID acquired in step 90. Then, of the processing parameters that have been read out, the cutoff frequency of the high-pass filter and the cutoff frequency of the low-pass filter in the second filter process are set in the second filter unit 68. Note that the above vehicle model ID is an example of second information.

In addition, in step 94, the parameter setting unit 74 acquires constitutional information of the driver 20 from the wearable terminal 52. In step 96, the parameter setting unit 74 reads out a threshold value registered in the threshold DB 44 in association with the constitutional information acquired in step 94, and sets the read-out threshold value in the abnormal value reduction unit 62. Note that the constitutional information is an example of the first information. In addition, in the EEG signal processing according to the second embodiment (FIG. 9), steps 94 and 95 may be executed before steps 90 and 92. In addition, steps 90 and 92 may be executed before step 112, and do not have to be executed at the beginning of the EEG signal processing (FIG. 9).

As described above, in the second embodiment, first information (physical information about the driver) related to artifacts caused by the subject's body movements contained in the subject's EEG signal is acquired, and at least the processing parameters of the abnormal value reduction process are set based on the acquired first information. This makes it possible to appropriately reduce artifacts caused by body movements.

In addition, in the second embodiment, second information (vehicle type ID) related to artifacts caused by the environment contained in the subject's EEG signal is acquired, and the processing parameters of the high-pass filter and low-pass filter of the second filter processing are set based on the acquired second information. This makes it possible to appropriately reduce artifacts caused by the environment.

In the second embodiment, the first information (physical information of the driver) related to artifacts caused by the subject's body movements is acquired, and the processing parameters (thresholds) of the abnormal value reduction processing are set based on the acquired first information. However, the present disclosure is not limited to this, and the processing parameters of the first filter processing (e.g., the interval of the moving average filter) may also be set in addition to the processing parameters of the abnormal value reduction processing based on the acquired first information. However, if the value of the interval of the moving average filter is made too long (e.g., 30 seconds or more), it may adversely affect the consciousness state determination processing in the subsequent stage, so it is preferable to set a value within the range of approximately 5 to 10 seconds.

In the second embodiment, the wearable terminal 52 detects the physical characteristics of the driver 20, such as the level of sweatiness, but the present disclosure is not limited to this. For example, a biosensor may be built into the seat body 12 of the vehicle seat 10, and the physical characteristics of the driver 20, such as the level of sweatiness, may be detected by the biosensor.

In addition, while the above describes an embodiment in which a moving average filter is applied as an example of a smoothing filter, the present disclosure is not limited to this, and other smoothing filters, such as a weighted average filter, a weighted moving average filter, or an exponential moving average filter, may also be applied.

In addition, although the above describes an embodiment in which the EEG signal processing ECU 26 of the vehicle seat 10 also performs the significant band determination process and the consciousness state determination process, the present disclosure is not limited to this. For example, the EEG signal processing ECU 26 of the vehicle seat 10 may output an EEG signal in which various artifacts have been reduced to an external device (e.g., the vehicle-side ECU 50), and the significant band determination process and the consciousness state determination process may be performed by the external device.

Furthermore, in the above description, an embodiment has been described in which an EEG signal processing ECU 26, which is an example of an EEG signal processing device according to the present disclosure, is incorporated in a vehicle seat 10 and processes the EEG signal of a driver 20 who drives the vehicle, but the present disclosure is not limited to this. According to the present disclosure, it is possible to obtain an EEG signal with reduced artifacts even when the subject is not in a resting state, and therefore the present disclosure is applicable to various EEG application technologies.

In addition, although the above describes a mode in which the EEG signal processing program 40 according to the present disclosure is pre-stored (installed) in the storage unit 32, the EEG signal processing program according to the present disclosure can also be provided in a form in which it is recorded on a non-temporary recording medium such as an HDD, SSD, or DVD.

Next, we will explain the experiments conducted by the inventors of the present application to verify the effectiveness of the EEG signal processing method disclosed herein.

In this experiment, a 15 Hz sine wave (see FIG. 10) was used as a signal simulating a true EEG signal. FIG. 11 shows the power spectrum obtained by performing FFT on this 15 Hz sine wave. In addition, in order to simulate various artifacts superimposed on the EEG signal, a pseudo-EEG signal was created by multiplying the 15 Hz sine wave by a 1 Hz sine wave that simulates communication artifacts and a 50 Hz sine wave that simulates environmental artifacts, and further adding random noise to simulate artifacts due to body movement (FIG. 12). FIG. 13 shows the power spectrum obtained by performing FFT on this pseudo-EEG signal. Note that the reason artifacts due to body movement are simulated by random noise is to confirm whether artifacts due to body movement can be reduced even if they are artifacts due to unexpected body movement.

In the power spectrum of the pseudo-EEG signal shown in Figure 13, frequency components corresponding to communication artifacts appear around 1 Hz, and frequency components corresponding to environmental artifacts appear around 50 Hz. In addition, random noise components corresponding to body movement artifacts appear in all frequency bands, and the 15 Hz frequency component corresponding to the true EEG signal is hidden by the noise.

As shown in FIG. 4, EEG signal processing using the EEG signal processing method according to the present disclosure involves acquiring an EEG signal and then performing the following processes in numerical order: (1) outlier reduction process (steps 102-106), (2) first filter process (moving average filter) (step 108), (3) DC offset reduction process (step 110), and (4) second filter process (high-pass filter and low-pass filter) (step 112). First, we will explain the results of an experiment in which each of the processes (1)-(4) was performed individually on the aforementioned pseudo-EEG signal.

Figure 14 shows the power spectrum of the EEG signal when only the (1) outlier reduction process is performed on the pseudo-EEG signal. (1) When only the outlier reduction process is performed, the 15 Hz frequency component, which corresponds to the true EEG signal, is attenuated. In addition, the peak at 1 Hz is high, and there is also a peak at 50 Hz with the same power as 15 Hz. This may result in an erroneous determination of the driver's consciousness in the subsequent stage.

Figure 15 shows the power spectrum of the EEG signal when only the (2) first filter process (moving average filter) is performed on the pseudo-EEG signal. When only the (2) first filter process (moving average filter) is performed, there is no attenuation of the 15 Hz frequency component that corresponds to the true EEG signal, but the peak at 1 Hz is high and there is also a peak at 50 Hz with the same power as 15 Hz. This may result in an erroneous determination of the driver's consciousness in the subsequent stage.

Figure 16 shows the power spectrum of the EEG signal when only the (3) DC offset reduction process is performed on the pseudo-EEG signal. When only the (3) DC offset reduction process is performed, the 1 Hz peak can be cut, but there is a peak at 50 Hz that has the same power as that at 15 Hz. This may result in an erroneous determination of the driver's consciousness state in the subsequent stage.

Figure 17 shows the power spectrum of the EEG signal when only the (4) second filtering process (high-pass filter and low-pass filter) is performed on the pseudo-EEG signal. When only the (4) second filtering process (high-pass filter and low-pass filter) is performed, a peak appears at approximately 10 Hz, and the 15 Hz frequency component corresponding to the true EEG signal is attenuated and hidden by other frequency components. This causes an erroneous determination of the driver's consciousness state in the subsequent stage.

As described above, in the power spectra of the EEG signals when each of the processes (1) to (4) is performed alone (Figures 14 to 17), the 15 Hz frequency component that simulates the true EEG signal is not significantly larger than the components in the other frequency bands, and it is clear that the performance in reducing artifacts is insufficient.

Next, we will explain the results when the pseudo-EEG signal is subjected to the processes (1) to (4) in the numerical order (in accordance with the EEG signal processing method of the present disclosure).

First, when the pseudo-EEG signal is subjected to the (1) outlier reduction process, the waveform of the EEG signal changes as shown in Figures 12 to 18, and its power spectrum changes as shown in Figures 13 to 19. In the power spectrum shown in Figure 19, the 50 Hz frequency component has decreased and the 15 Hz frequency component has become slightly more noticeable, but the artifacts have not been sufficiently reduced, so it can be said to be in an incomplete state. For this reason, there is a risk of an erroneous determination being made as to which frequency bands are significant when determining the driver's state of consciousness in a later stage.

Next, when (2) the first filter process (moving average filter) is performed on the EEG signal that has been subjected to (1) the abnormal value reduction process, the waveform of the EEG signal changes as shown in FIG. 18 to FIG. 20, and its power spectrum changes as shown in FIG. 19 to FIG. 21. In the power spectrum shown in FIG. 21, the 15 Hz frequency component becomes slightly more noticeable, but the artifacts have not been sufficiently reduced, so it can be said to be in an incomplete state. For this reason, there is a risk of an erroneous determination being made as to which frequency band is significant when determining the driver's state of consciousness in a later stage.

Next, when the EEG signal that has been subjected to (1) anomaly reduction processing and (2) first filter processing (moving average filter) is subjected to (3) DC offset reduction processing, the waveform of the EEG signal changes as shown in FIG. 20 to FIG. 22, and its power spectrum changes as shown in FIG. 21 to FIG. 23. As is clear from comparing FIG. 22 with FIG. 20, the EEG signal that has been subjected to DC offset reduction processing has its center of amplitude shifted to near 0. In the power spectrum shown in FIG. 23, although the frequency components (artifacts during communication) around 1 Hz have been suppressed to about half, each artifact has not been sufficiently reduced, so it can be said to be in an incomplete state. For this reason, there is a risk of an erroneous determination being made as to which frequency band is significant in the determination of the driver's consciousness state in the subsequent stage.

Furthermore, when the EEG signal that has been subjected to (1) outlier reduction processing → (2) first filtering processing (moving average filter) → (3) DC offset reduction processing is subjected to (4) second filtering processing (high-pass filter and low-pass filter), the waveform of the EEG signal changes as shown in Figures 22 to 24, and its power spectrum changes as shown in Figures 23 to 25. In the power spectrum shown in Figure 25, each artifact is sufficiently reduced, and the frequency component of 15 Hz, which corresponds to the true EEG signal, shows a prominent peak. Therefore, in the determination of the driver's consciousness state in the subsequent stage, it can be determined without fail that the frequency band around 15 Hz is significant.

Furthermore, in order to verify the effectiveness of performing each of the processes (1) to (4) in numerical order (performing them in the order that conforms to the EEG signal processing method of the present disclosure), the inventors of the present application conducted an experiment in which the execution order of each process was changed and the power spectrum was confirmed. The results of this experiment will now be described.

The first execution order is to perform the DC offset reduction process first. That is, the pseudo-EEG signal is subjected to the following processes in the order: (3) DC offset reduction process → (1) abnormal value reduction process → (2) first filter process (moving average filter) → (4) second filter process (high-pass filter and low-pass filter). The power spectrum of the EEG signal obtained by performing each process in the first execution order is shown in FIG. 26.

In the power spectrum shown in FIG. 26, the frequency component at 15 Hz, which corresponds to the true EEG signal, has the maximum peak, but the peak of the frequency component near 25 Hz is no less high. Therefore, if a significant band determination process is performed on the power spectrum shown in FIG. 26 to determine that the frequency band in which the maximum peak occurs is significant, there is a risk of misjudging the significant frequency band. Furthermore, the power spectrum shown in FIG. 26 has a high level on average for the frequency bands near 5 Hz to 28 Hz. Therefore, even if a significant band determination process is performed on the power spectrum shown in FIG. 26 to determine that the frequency band in which the result of integrating each frequency band shows the maximum value is significant, there is a risk of misjudging the significant frequency band.

The second execution order is to perform the second filter processing (high-pass filter and low-pass filter) first. That is, the pseudo-EEG signal is subjected to the following processing in the order of (4) second filter processing (high-pass filter and low-pass filter) → (1) outlier reduction processing → (2) first filter processing (moving average filter) → (3) DC offset reduction processing. The power spectrum of the EEG signal obtained by performing each processing in the second execution order is shown in Figure 27.

The power spectrum shown in Figure 27 has a maximum peak near 7 Hz, and the peak near 15 Hz, which corresponds to the true EEG signal, is attenuated. For this reason, if a significant band determination process is performed on the power spectrum shown in Figure 27 to determine that the frequency band in which the maximum peak occurs is significant, an erroneous determination of the significant frequency band will occur. Furthermore, the power spectrum shown in Figure 27 has several peaks in the frequency band near 7 Hz to 28 Hz. For this reason, even if a significant band determination process is performed on the power spectrum shown in Figure 27 to determine that the frequency band in which the result of integrating each frequency band shows a maximum value is significant, there is a risk of an erroneous determination of the significant frequency band.

The third execution order is to perform the first filter processing (moving average filter) first. That is, the pseudo-EEG signal is subjected to the following processing in the order of (2) first filter processing (moving average filter) → (1) outlier reduction processing → (3) DC offset reduction processing → (4) second filter processing (high-pass filter and low-pass filter). The power spectrum of the EEG signal obtained by performing each processing in the third execution order is shown in Figure 28.

In the power spectrum shown in Figure 28, the peak near 15 Hz, which corresponds to the true EEG signal, has been cut off and disappears, and there are high peaks near 8 Hz and 23 Hz. For this reason, when a significant band determination process is performed on the power spectrum shown in Figure 28 to determine that the frequency band in which the maximum peak occurs is significant, an erroneous determination of the significant frequency band occurs. Also, when a significant band determination process is performed to determine that the frequency band in which the result of integrating each frequency band shows the maximum value is significant, there is a risk of an erroneous determination of the significant frequency band.

The fourth execution order is to perform the DC offset reduction process last. That is, the pseudo-EEG signal is subjected to the following processes in the order: (1) outlier reduction process → (2) first filter process (moving average filter) → (4) second filter process (high-pass filter and low-pass filter) → (3) DC offset reduction process. The power spectrum of the EEG signal obtained by performing each process in the fourth execution order is shown in FIG. 29.

The power spectrum shown in Figure 29 has high peaks near 5 Hz and 7 Hz, and no clear peak is present near 15 Hz, which corresponds to the true EEG signal. For this reason, when a significant band determination process is performed on the power spectrum shown in Figure 29 to determine that the frequency band in which the maximum peak occurs is significant, an erroneous determination of the significant frequency band occurs. Also, when a significant band determination process is performed to determine that the frequency band in which the result of integrating each frequency band shows the maximum value is significant, there is a risk of an erroneous determination of the significant frequency band.

The fifth execution order is to perform the outlier reduction process last. That is, the pseudo-EEG signal is subjected to the following processes in the order: (2) first filter process (moving average filter) → (3) DC offset reduction process → (4) second filter process (high-pass filter and low-pass filter) → (1) outlier reduction process. The power spectrum of the EEG signal obtained by performing each process in the fifth execution order is shown in Figure 30.

The power spectrum shown in Figure 30 has a high peak near 11 Hz, and although there is a peak near 15 Hz, which corresponds to the true EEG signal, its height is relatively low. For this reason, when a significant band determination process is performed on the power spectrum shown in Figure 30 to determine that the frequency band in which the maximum peak occurs is significant, an erroneous determination of the significant frequency band occurs. Also, when a significant band determination process is performed to determine that the frequency band in which the result of integrating each frequency band shows the maximum value is significant, there is a risk of an erroneous determination of the significant frequency band.

The sixth execution order is to perform the first filter process (moving average filter) last. That is, the pseudo-EEG signal is subjected to the following processes in the order: (1) outlier reduction process → (3) DC offset reduction process → (4) second filter process (high-pass filter and low-pass filter) → (2) first filter process (moving average filter). The power spectrum of the EEG signal obtained by performing each process in the sixth execution order is shown in FIG. 31.

In the power spectrum shown in Figure 31, there is a high peak near 27 Hz, and the next highest peak is near 20 Hz. There is no clear peak near 15 Hz, which corresponds to the true EEG signal. For this reason, if a significant band determination process is performed on the power spectrum shown in Figure 31 to determine that the frequency band in which the maximum peak occurs is significant, an erroneous determination of the significant frequency band will occur. Also, if a significant band determination process is performed to determine that the frequency band in which the result of integrating each frequency band shows the maximum value is significant, there is a risk of an erroneous determination of the significant frequency band.

The seventh execution order reverses the order of execution of the DC offset reduction process and the first filter process (moving average filter). That is, the pseudo-EEG signal is subjected to the following processing in the order of (1) abnormal value reduction process → (3) DC offset reduction process → (2) first filter process (moving average filter) → (4) second filter process (high-pass filter and low-pass filter). The power spectrum of the EEG signal obtained by performing each process in the seventh execution order is shown in FIG. 32.

In the power spectrum shown in Figure 32, there is a high peak near 22 Hz, and the next highest peak is near 19 Hz. Although there is a small peak near 15 Hz, which corresponds to the true EEG signal, the height of the peak is low. For this reason, when a significant band determination process is performed on the power spectrum shown in Figure 32, in which the frequency band in which the maximum peak occurs is determined to be significant, an erroneous determination of the significant frequency band occurs. Also, when a significant band determination process is performed in which the frequency band in which the result of integrating each frequency band shows the maximum value is determined to be significant, there is a risk of an erroneous determination of the significant frequency band.

From the above results, it has been confirmed that performing each of the processes (1) to (4) on an EEG signal in numerical order (performing them in the order that conforms to the EEG signal processing method of the present disclosure) is extremely effective in obtaining an EEG signal that is close to a true EEG signal with various artifacts reduced.

10 Vehicle seat 12 Seat body 24 Electroencephalogram measuring device 26 Electroencephalogram signal processing ECU
40 EEG signal processing program 60 Acquisition unit 62 Abnormal value reduction unit 64 First filter unit 66 DC offset reduction unit 68 Second filter unit 70 Significant band determination unit 72 Consciousness state determination unit 74 Parameter setting unit

Claims (9)

an environmental artifact added to an electroencephalogram signal of a subject who is driving a vehicle is measured in advance for each vehicle model, and in accordance with a measurement result of the artifact, the vehicle model as second information related to the environmental artifact included in the electroencephalogram signal of the subject is stored in a storage unit in association with the first frequency and the second frequency in a second filtering process in which the electroencephalogram signal is filtered through a high-pass filter having a cutoff frequency of a first frequency and a low-pass filter having a cutoff frequency of a second frequency higher than the first frequency, for each vehicle model;
Acquiring an electroencephalogram signal of a subject who is driving a vehicle and acquiring the type of vehicle that the subject is driving;
An abnormal value reduction process is performed on the acquired electroencephalogram signal to reduce abnormal values having an amplitude equal to or greater than a threshold value.
performing a first filtering process in which a smoothing filter is applied to the electroencephalogram signal that has been subjected to the abnormal value reduction process;
performing a DC offset reduction process for reducing a DC offset component on the electroencephalogram signal that has been subjected to the first filtering process;
A second filtering process is performed on the electroencephalogram signal that has been subjected to the DC offset reduction process, in which the first frequency that is stored in the storage unit in association with the vehicle model of the acquired vehicle is set as a cutoff frequency of the high-pass filter, and the second frequency that is stored in the storage unit in association with the vehicle model of the acquired vehicle is set as a cutoff frequency of the low-pass filter.
EEG signal processing method. 2. The electroencephalogram signal processing method according to claim 1, wherein the abnormal value reduction process comprises replacing a data value whose amplitude is equal to or greater than a threshold value with an intermediate value between the data before and after the data value. determining a significant frequency band in a power spectrum of the electroencephalogram signal that has been subjected to the second filtering process;
3. The electroencephalogram signal processing method according to claim 1, further comprising the step of determining a state of consciousness of the subject from the determined significant frequency band. 4. The electroencephalogram signal processing method according to claim 1, wherein an electroencephalogram signal is acquired from a subject in a non-resting state. 5. The electroencephalogram signal processing method according to claim 1, further comprising acquiring an electroencephalogram signal from a subject who is driving a vehicle. The electroencephalogram signal processing method according to any one of claims 1 to 5, wherein the threshold value in the abnormal value reduction process is 150 μV or a value exceeding this. a storage unit that measures in advance for each vehicle type an artifact caused by an environment added to an electroencephalogram signal of a subject who is driving a vehicle, and stores, according to a measurement result of the artifact, the vehicle type as second information related to the artifact caused by the environment included in the electroencephalogram signal of the subject, in association with the first frequency and the second frequency in a second filtering process that filters the electroencephalogram signal through a high-pass filter having a cutoff frequency of a first frequency and a low-pass filter having a cutoff frequency of a second frequency higher than the first frequency, for each vehicle type;
an acquisition unit that acquires an electroencephalogram signal of a subject who is driving a vehicle and acquires a type of the vehicle that the subject is driving;
an abnormal value reduction unit that performs an abnormal value reduction process for reducing abnormal values having amplitudes equal to or greater than a threshold value for the electroencephalogram signal acquired by the acquisition unit;
a first filter unit that performs a first filter process of applying a smoothing filter to the electroencephalogram signal that has been subjected to the abnormal value reduction process by the abnormal value reduction unit;
a DC offset reduction unit that performs a DC offset reduction process to reduce a DC offset component on the electroencephalogram signal that has been subjected to the first filtering process by the first filter unit;
a second filter unit that performs a second filter process on the electroencephalogram signal that has been subjected to the DC offset reduction process by the DC offset reduction unit, the second filter process setting the first frequency that is stored in the storage unit in association with the vehicle model acquired by the acquisition unit as a cutoff frequency of the high-pass filter, and setting the second frequency that is stored in the storage unit in association with the vehicle model as a cutoff frequency of the low-pass filter;
13. An electroencephalogram signal processing device comprising: A seat body on which a driver of the vehicle sits;
an electroencephalogram (EEG) measuring unit that measures an electroencephalogram of a driver of a vehicle seated in the seat body and outputs an electroencephalogram signal;
8. The electroencephalogram signal processing apparatus according to claim 7, further comprising:
1. A vehicle seat comprising: an environmental artifact added to an electroencephalogram signal of a subject who is driving a vehicle is measured in advance for each vehicle model, and in accordance with a measurement result of the artifact, the vehicle model as second information related to the environmental artifact included in the electroencephalogram signal of the subject is stored in a storage unit in association with the first frequency and the second frequency in a second filtering process in which the electroencephalogram signal is filtered through a high-pass filter having a cutoff frequency of a first frequency and a low-pass filter having a cutoff frequency of a second frequency higher than the first frequency, for each vehicle model;
On the computer,
Acquiring an electroencephalogram signal of a subject who is driving a vehicle and acquiring the type of vehicle that the subject is driving;
Acquire an electroencephalogram (EEG) signal from a subject who is driving a vehicle;
An abnormal value reduction process is performed on the acquired electroencephalogram signal to reduce abnormal values having an amplitude equal to or greater than a threshold value.
performing a first filtering process in which a smoothing filter is applied to the electroencephalogram signal that has been subjected to the abnormal value reduction process;
performing a DC offset reduction process for reducing a DC offset component on the electroencephalogram signal that has been subjected to the first filtering process;
A second filtering process is performed on the electroencephalogram signal that has been subjected to the DC offset reduction process, in which the first frequency that is stored in the storage unit in association with the vehicle model of the acquired vehicle is set as a cutoff frequency of the high-pass filter, and the second frequency that is stored in the storage unit in association with the vehicle model of the acquired vehicle is set as a cutoff frequency of the low-pass filter.
The electroencephalogram signal processing program for executing processing including the above.