JP6821549B2 - Biological signal processing device, program and method for determining the generation of biological signals based on the representative values of acceleration components - Google Patents

Description

The present invention relates to a technique for detecting a human biological signal.

In recent years, techniques have been developed in which various biological signals caused by various activities of humans and animals are detected by sensors and the biological data obtained by signal processing is used in various situations. Here, as the sensor, for example, a wristwatch type pulse sensor, an earphone type pulse sensor, a headband type brain wave sensor, or the like is used. In addition, the biological signal detected by such a sensor is processed and processed by, for example, a smartphone carried by a user, and is used in various applications.

As a specific technical example using such a sensor, the inventor of the present application has invented a device provided with a myoelectric sensor that detects a myoelectric signal as a biological signal and capable of identifying facial expressions such as smiles and clenching. (See Patent Document 1). Here, the reference electrode and the detection electrode of this myoelectric sensor are arranged so as to be in contact with the skin surface at some point near the cheek from around the left and right (or right and left) auricles, respectively. ..

In addition, as myoelectric signal processing, this device has a power value VLF in the first frequency band related to artifacts (noise signals other than the target signal) and power in the second frequency band related to the signal at the time of the first facial expression. When the value LF is calculated and the first power VLF is equal to or less than the first reference power VLFBase and the second power LF is larger than the second reference power LFBase, at the time of the first facial expression. It is determined that there is.

Further, the inventor of the present application discloses in Patent Document 2 a method of obtaining a distance by frequency warping as a similarity in order to compare the power feature amount of the frequency axis of the myoelectric signal with the reference feature amount.

JP-A-2017-0292323 JP-A-2017-140198

As described above, the inventor of the present application has been researching and developing various biological signal processing methods, and in particular, has tackled the problem of determining the generation of biological signals without analyzing the frequency of the signal.

For example, in the device disclosed in the above-mentioned Patent Document 1, it is necessary to measure the reference power for a specific frequency in advance, and in the method disclosed in the above-mentioned Patent Document 2, the power with respect to the frequency is used as a feature vector (feature amount vector). It was necessary to find the similarity as the shape of the mountain of the spectrum). On the other hand, by determining whether or not a biological signal is generated without performing frequency analysis, it becomes easier to realize processing with a smaller amount of calculation.

Furthermore, the inventor of the present application has recognized that it is also an important issue to realize a process that is robust against large-amplitude artifacts mixed in by blinking, etc., while conducting various experiments on biological signal determination processing. It was. By the way, it is known that this artifact is generated when the electrodes are displaced due to blinking or the like.

Therefore, an object of the present invention is to provide a biological signal processing device, a program, and a method that do not require frequency analysis in determining the generation of biological signals and are robust against the generation of noise due to electrode misalignment.

According to the present invention, it is a biological signal processing device that processes an input signal that may include a myoelectric signal.
Acceleration component generation means that applies second-order difference filter processing to the input signal to generate acceleration component data as the second-order difference of the input signal, and
A monotonous decrease in the value indicating the degree of bias of the data in the predetermined time interval in the acceleration component data and the length of the time interval in which the acceleration component of the acceleration component data in the predetermined time interval remains continuously within the predetermined range. A weight that serves as a function is calculated, and a value indicating the degree of bias of the data is weighted by the weight to calculate a representative value related to the degree of bias of the data, or the predetermined time interval in the acceleration component data. A representative value calculation means in which a value indicating the degree of bias of the data in the above is used as a representative value related to the degree of bias of the data.
When the representative value exceeds a predetermined threshold value, a biological signal processing device including a signal generation determination means for determining that the myoelectric signal has been generated is provided.

Further, as an embodiment of the biometric signal processing device according to the present invention, the biometric signal processing device includes band-stop filter processing for reducing noise related to a commercial power source for the input signal before generating the acceleration component data. It is also preferable to further have a pre-filter processing means for performing a low-pass filter processing for extracting a bias fluctuation component.

Further, in the signal generation determination means of the biological signal processing device according to the present invention, the time-series data of the representative value moves upward from the lower side across the line of the first threshold value with the passage of time, and then the first threshold value is obtained. It is also preferable to determine that the biological signal has been generated n times when the hysteresis as a general behavior that crosses the line of the second threshold value smaller than the upper side and goes downward is n degrees (n is a natural number). ..

Further, as another embodiment of the biological signal processing device according to the present invention, the biological signal processing device calculates the average power frequency of the input signal in the time interval related to the determination that the myoelectric signal has been generated. It is also preferable to further have a biological signal discriminating means for determining the type of the generated myoelectric signal based on the height of the average power frequency.

Further, as another embodiment of the above-mentioned biological signal discrimination, the present biological signal processing apparatus defines an amount including the standard deviation and the average power frequency of the input signal as the feature amount related to the generation of the myoelectric signal. Above,
A state and muscle in which a certain type of myoelectric signal is generated based on any one of the MT (Mahalanobis Taguchi) method, the MTA (Mahalanobis-Taguchi Adjoint) method, the T method, and the RT (Recognition Taguchi) method. Depending on the feature amount related to the input signal in the state where no electric signal is generated and no artifact as noise, which is a signal other than the certain type of myoelectric signal, is generated . A first unit space is set, a second unit space is set according to the feature amount related to the input signal in a state where no myoelectric signal is generated, and a certain type of myoelectric signal is further set. A third unit space is set according to the feature amount related to the input signal in the generated state, and the third unit space is set.
In the time interval related to the determination that some myoelectric signal has been generated, the feature amount related to the input signal is calculated, and the calculated feature amount is used in the first unit space, the second unit space, and the third unit space. The first degree of separation, the second degree of separation, and the third degree of separation, which are the degrees of separation from each of the unit spaces of, are calculated.
When the amount obtained by subtracting the first degree of separation and the third degree of separation from the second degree of separation exceeds a predetermined threshold value, it is considered that some myoelectric signal is a certain type of myoelectric signal. It is also preferable to further have a biological signal discriminating means for determining.

Further, in the biometric signal processing device of the present invention, the myoelectric signal is a device attached to the user's head, and the reference electrode is one skin from around the auricle to the vicinity of the cheek on the left (or right). It is also preferable that the signal is acquired by a device having an electrode configuration that is in contact with a position and the detection electrode is in contact with one skin position near the cheek from the area around the right (or left) pinna. Further, the myoelectric signal is also preferably a myoelectric signal due to muscle of the corner of the mouth up.

According to the present invention, it is also a program that functions a computer mounted on a device that processes an input signal that may include a myoelectric signal.
Acceleration component generation means that applies second-order difference filter processing to the input signal to generate acceleration component data as the second-order difference of the input signal, and
A monotonous decrease in the value indicating the degree of bias of the data in the predetermined time interval in the acceleration component data and the length of the time interval in which the acceleration component of the acceleration component data in the predetermined time interval remains continuously within the predetermined range. A weight that serves as a function is calculated, and a value indicating the degree of bias of the data is weighted by the weight to calculate a representative value related to the degree of bias of the data, or the predetermined time interval in the acceleration component data. A representative value calculation means in which a value indicating the degree of bias of the data in the above is used as a representative value related to the degree of bias of the data.
When the representative value exceeds a predetermined threshold value, a biometric signal processing program that causes a computer to function as a signal generation determination means for determining that the myoelectric signal has been generated is provided.

According to the present invention, further, it is a biological signal processing method by a computer mounted on a device for processing an input signal that may include a myoelectric signal.
A step of performing a second-order difference filter processing on the input signal to generate acceleration component data as a second-order difference of the input signal, and
A monotonous decrease in the value indicating the degree of bias of the data in the predetermined time interval in the acceleration component data and the length of the time interval in which the acceleration component of the acceleration component data in the predetermined time interval remains continuously within the predetermined range. A weight that serves as a function is calculated, and a value indicating the degree of bias of the data is weighted by the weight to calculate a representative value related to the degree of bias of the data, or the predetermined time interval in the acceleration component data. The step of using a value indicating the degree of bias of the data in the above as a representative value related to the degree of bias of the data, and
When the representative value exceeds a predetermined threshold value, a biological signal processing method including a step of determining that the myoelectric signal has been generated is provided.

According to the biological signal processing apparatus, program and method of the present invention, frequency analysis is not required in determining the generation of biological signals, and biological signal processing that is robust against the generation of noise due to electrode misalignment is realized.

It is a schematic diagram which shows one Embodiment of the biological signal processing apparatus by this invention. It is a graph for demonstrating the Example which carried out the biological signal processing which concerns on this invention. It is a graph for demonstrating the Example which carried out the biological signal processing which concerns on this invention. It is a graph for demonstrating the weight used for calculating the representative value which concerns on this invention. It is a graph which shows one Embodiment of the signal generation determination processing using the hysteresis of a typical value. It is a flowchart which shows the outline of one Embodiment of the biological signal processing method by this invention. It is a schematic diagram which shows the other embodiment of the biological signal processing apparatus by this invention. It is a schematic diagram which shows the further other embodiment of the biological signal processing apparatus by this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[Biosignal processing device]
FIG. 1 is a schematic view showing an embodiment of a biological signal processing device according to the present invention.

FIG. 1 shows glasses 1 with an myoelectric sensor as an embodiment of the biological signal processing device according to the present invention. The glasses 1 with an myoelectric sensor is a glasses-type device that can be attached to the head of a living body (for example, a human user) to acquire a biological signal. In the present embodiment, the biological signal acquired by this device is a "myoelectric signal" as an electric signal generated due to the movement of a portion in the face or the movement related to a facial expression. In addition, the acquired biological signal may include a "(noise) signal caused by electrode misalignment" or the like generated by such movement.

Here, as the movement related to the movement of the intrafaceal part to be detected or the movement related to the facial expression, for example, at least one of (smiling) raising the corner of the mouth, clenching (or clenching), chewing, and blinking (blinking motion). One can be set.

Similarly, as shown in FIG. 1, the glasses 1 with an electromyographic sensor
(A) A frame portion as a main body of the device provided with a signal processing box 11 which is a portion for capturing and processing biological signals, and
(B) A positive electrode pad 13 and a negative electrode portion as an electrode portion for receiving a biological signal, which are arranged at a position in contact with the skin of the head and capable of receiving at least a part of the weight of the frame portion 11. Electrode pad 14 and
(C) An elastic support portion provided with a conductive path for transmitting a biological signal received via the positive electrode pad 13 and the negative electrode pad 14 to the signal processing box 11.
(D) It has a nose pad electrode portion 15 which is arranged at a position in contact with the vicinity of the upper part of the nose and is provided with a ground (GND) electrode or a noise canceling electrode when receiving a biological signal. The noise canceling electrode (d) may be a DRL (Driven Right Leg) electrode that reduces common mode noise caused by a commercial power source or the like.

Further, the positive electrode pad 13 of the above (b) functions as a detection electrode or a positive electrode when receiving a biological signal, while the negative electrode pad 14 functions as a reference electrode or a negative electrode when receiving a biological signal. .. The biological signal is detected and acquired as a potential difference between the positive electrode pad 13 and the negative electrode pad 14.

As described above, in the glasses 1 with the myoelectric sensor, the electrode portions such as the positive electrode pad 13 and the negative electrode pad 14 function not only as a means for receiving the biological signal but also as a means for supporting the device main body portion. .. Further, the elastic support portion elastically connects these "electrode portions" to the device main body portion. As a result, for example, even if the attached head moves significantly, these electrode portions are stabilized in the vicinity of a predetermined position on the skin of the head with the weight of the device main body portion transmitted through the elastic portion called the elastic support portion. It becomes possible to keep in contact with each other.

Here, one mounting example will be described. The human cheekbones project laterally when viewed from the front of the face, but if the positive electrode pads 13 and the negative electrode pads 14 are brought into contact with the skin slightly above the widest part of the cheekbones, the left and right sides Since the distance between the "electrode portions" is narrower than the maximum width of the cheekbone and is caught in the widened portion of the upper part of the cheekbone, the glasses 1 with an myoelectric sensor are stably supported.

Further, although not shown, the glasses 1 with a myoelectric sensor is provided with a conductive path for taking a biological signal from the electrode portion into the signal processing box 11, and the biological signal received by the electrode portion is sent to the signal processing box. It enables stable and reliable uptake to 11. That is, the conductive path functions as a stable electrical transmission path that connects the left and right signal processing boxes 11 and each electrode portion.

As a modification, the modern part of the glasses 1 may have a function as a GND electrode or a noise canceling electrode. In this case, the nasal pad electrode portion 15 can be omitted to make the nasal padless. Further, as a further modification, the electrodes of the modern portion and the nasal pad electrode portion 15 may be electrically conductive, and the plurality of electrodes may function as GND electrodes.

Further, as a matter of course, the positive electrode pad 13 and the negative electrode pad 14 may be in an interchangeable shape connected to the left temple portion and the right temple portion, respectively. In any case, by arranging the positive electrode pad 13 and the negative electrode pad 14 separately on the left and right, it is possible to capture the two left and right muscle activities. The muscular activity that creates the facial expression "smile" generally occurs simultaneously on both sides, not on either side. Therefore, rather than arranging one set of electrodes that make up one channel on either the left or right side, it is better to arrange each of the electrodes that make up one set separately on the left and right to capture the entire left and right muscle activity. Therefore, in the end, a more stable and large myoelectric signal can be obtained.

In addition, by separating the left and right electrodes in this way, various myoelectric signals caused by muscle activity in the head such as left and right eye movements and tooth clenching (clenching), and even the mouth. It is also possible to more reliably capture various signals such as changes in contact resistance between the skin and electrodes caused by irregularities on the skin surface caused by opening and closing of the skin and chewing.

Further, two elastic support portions are provided in the present embodiment, and the positive electrode pad 13 and the negative electrode pad 14, respectively, are located at the position of the skin below the temple and when the face is viewed from the front. It elastically presses (contacts) the skin slightly above the widest part of the cheekbone, thereby functioning as a support structure for supporting the glasses 1 with an myoelectric sensor. Further, both the positive electrode pad 13 and the negative electrode pad 14 can come into contact with the skin at any position within the range from the upper cheek to the base of the ear via the temple, and the frame of the elastic support portion. The position with respect to the part is adjusted.

Similarly, as shown in FIG. 1, in the present embodiment, the signal processing box 11 arranged in the temple portion on the right side has a built-in battery for driving the processing unit, while the signal arranged in the temple portion on the left side. The processing box 11 includes a biological signal processing unit 12 that processes the acquired biological signal with the power supplied from the battery. It is also preferable that the weights of the left and right signal processing boxes 11 are set to be substantially (almost) equal. As a result, it is possible to balance the left and right sides of the weight of the glasses 1 with the myoelectric sensor, and it is possible to realize a good wearing feeling without bias.

By the way, these batteries and the biological signal processing unit 12 can be built in the frame unit instead of the box, and the entire glasses 1 with the myoelectric sensor can be designed to be the same in appearance as ordinary glasses. Such compactification of the processing unit is possible by using the biological signal processing unit 12 of the embodiment, which will be described in detail later.

Further, as described above, when the positive electrode pad 13 and the negative electrode pad 14 are in contact with the skin at a position within the range from the upper cheek to the base of the ear via the temple, the biological signal that can be acquired is myoelectric telegraph. It is not limited to the issue. For example, in addition to biological signals based on biopotentials such as electrooculogram signals and brain waves that can be detected from a position near the ear, sensor devices other than biopotential sensors are required, but signals related to body temperature and sweating, pulse waves, etc. It is also possible to detect and acquire.

[Structure of biological signal processing unit]
Similarly, according to FIG. 1, the biological signal processing unit 12 included in the signal processing box 11 is an input signal processing unit that can include a biological signal.
(A) Acceleration component generation unit 123 that generates "acceleration component data" of the input signal, and
(B) The representative value calculation unit 124 that calculates the "representative value" related to the degree of bias of the data in the predetermined time interval in the "acceleration component data", and
(C) It is characterized by having a signal generation determination unit 125 for determining the generation of a biological signal based on the size of the calculated "representative value". Here, for example, the standard deviation SD may be used as the “representative value”, but more preferably, the value SD _W weighted with the standard deviation SD, which will be described in detail later, can be adopted.

In this way, the biological signal processing unit 12 does not perform frequency analysis that requires a large amount of calculation in determining whether or not a biological signal is generated, but rather a "representative value" in the "acceleration component data". Is calculated and the judgment process is performed. By applying such a process with less calculation load, the inventor of the present application is not affected by noise due to electrode misalignment, and more reliably determines whether or not the biological signal to be detected is generated. It was found that it can be judged.

In fact, a biological signal such as a myoelectric signal is an AC signal having a wide frequency component because it is generated due to the activity of a large number of cells, for example, unlike vibration caused by an artificial machine or the like. According to the biological signal processing unit 12, in determining the generation of such a biological signal, attention is paid to the "representative value" of the "acceleration component data", so that frequency analysis, which requires a large amount of calculation, becomes unnecessary. In addition, robust biological signal processing is realized even when noise is generated due to electrode misalignment.

Similarly, in the functional block diagram of the embodiment shown in FIG. 1, the biometric signal processing unit 12 includes a signal conversion unit 121, a notch filter unit 122a, and a low-pass filter (LPF, Low-pass filter) unit 122b. The functional components are an acceleration component generation unit 123 including a second-order difference filter unit 123a, a representative value calculation unit 124, a signal generation determination unit 125, and a biological signal determination unit 126.

Here, the biological signal processing unit 12 stores one embodiment of the biological signal processing program according to the present invention, and also has a computer function. By executing this biological signal processing program, the biological signal is processed. Perform the process. Further, the above-mentioned functional component can be regarded as a function of the biological signal processing program stored in the biological signal processing unit 12. Further, the processing flow shown by connecting the functional components of the biological signal processing unit 12 in FIG. 1 with arrows is also understood as an embodiment of the biological signal processing method according to the present invention. It is also preferable that the signal processing box 11 on the left side is further provided with a signal interface 127 together with the biological signal processing unit 12.

Similarly, in FIG. 1, the signal conversion unit 121 has a DRL circuit as a myoelectric sensor that reduces common mode noise caused by a commercial power source or the like. This DRL circuit
(A) A positive electrode (for detection) electrically connected to the positive electrode pad 13 and
(B) The AC component of the potential difference between the negative electrode pad 14 and the electrically connected negative (reference) electrode.
(C) Amplification is performed by differential amplification between the nasal pad electrode 15 and the GND potential of the electrically connected GND electrode, and this analog biological signal is digitized at a constant sampling frequency.

This makes it possible to detect the skin potential in the range of plus or minus 0.1 to several hundred μV, for example. Further, as a condition for this digitization, it is also preferable to perform analog / digital (A / D) conversion at a sampling frequency of 500 Hz or more and a quantization rate of 10 bits or more. It should be noted that such a circuit configuration can be realized by using, for example, TGAM1 manufactured by Neurosky.

The pre-filter processing unit 122 has a notch filter unit 122a and a low-pass filter (LPF) unit 122b. Of these, the notch filter unit 122a performs band-stop filter processing for reducing periodic noise related to a commercial power source (which is often mixed) with respect to the input signal before generating the acceleration component data. Incidentally, the above-mentioned TGAM1 manufactured by Neurosky is equipped with a notch filter that reduces noise derived from a commercial power source, and this can be used as the notch filter unit 122a.

On the other hand, the LPF unit 122b performs LPF processing for extracting a bias fluctuation component from the input signal subjected to the band removal filter processing. Specifically, the LPF unit 122b may perform LPF processing by performing a high-pass filter (HPF, High-Pass Filter) process on the input signal and subtracting the result from the original input signal. Here, for example, a DC blocker can be used as the HPF.

The DC blocker is a filter for removing the DC bias component (ultra-low frequency component) from the input signal and extracting the AC component. The following equation (1) y [n] = x [n] −x [n-1 ] ＋ r * y [n-1]
It works under the difference equation like. Here, n is the sample position (sample index), and x [n] and y [n] are the input signal and the output signal of the sample position n, respectively. Further, the coefficient r takes a value of 0 to 1, and when r = 0, this filter is equivalent to the difference filter described below. Incidentally, in the embodiment described later with reference to FIG. 2, r = 0.9 is set.

The acceleration component generation unit 123 has a second-order difference filter unit 123a, and generates acceleration component data of the input signal subjected to LPF processing. Specifically, the second-order difference filter unit 123a can be configured to perform the difference filter process twice on the input signal. The difference equation showing the principle of the difference filter used here is the following equation (2) y [n] ＝ x [n] −x [n-1]
It becomes the street. In the above equation (2), n is the sample position (sample index), and x [n] and y [n] are the input signal and the output signal of the sample position n, respectively.

In general, when a digital filter is used like the acceleration component generation unit 123, the more advanced the digital filter, the larger the amount of calculation. This increase in the amount of calculation causes a decrease in battery life in a mobile device such as the glasses 1 with a myoelectric sensor, which is a big problem. On the other hand, the acceleration component generation unit 123 does not use a filter including trigonometric functions, but uses a filter having a low order to process the biological signal, so that the increase in the amount of calculation which becomes a problem is suppressed. You can do it.

Similarly, in FIG. 1, the representative value calculation unit 124 divides the acceleration component data generated by the acceleration component generation unit 123 into predetermined time intervals (window analysis sections), and relates to the degree of data bias in each window analysis section. Calculate the representative value. In general, the time-series data output from the biosensor can be fed back to the user in real time via the user interface by sequentially analyzing it in real time, which makes it very easy to use. At this time, by providing a window analysis section in advance and performing sequential analysis while shifting the analysis section, it is possible to perform analysis processing in substantially real time.

In the present embodiment, when the sampling frequency of digitization in the signal conversion unit 121 is 512 Hz, the representative value calculation unit 124 sets the window analysis section to 128 samples and collects the time series data of the acceleration component every 0.25 seconds (every 128 samples). ), The standard deviation SD in the acceleration component data in the section is calculated for each section. As a modification, every time 64 samples of time series data of acceleration components are input, the standard deviation SD may be calculated using the most recently input 128 samples as the window analysis interval.

Naturally, the value calculated here is not limited to the standard deviation SD, and various values can be adopted as long as it is a value related to the degree of bias of the acceleration component data in the window analysis section.

The representative value calculation unit 124 also calculates a weight W that is a monotonically decreasing function for the length of the time interval (sample length len_th) in which the acceleration component in each window analysis interval remains continuously within a predetermined range. The weight W will be described in detail later with reference to the examples shown in FIGS. 2 and 4.

After that, the representative value calculation unit 124 determines the value obtained by weighting the calculated standard deviation SD with the similarly calculated weight W as the representative value SD _W. Specifically, the following equation (3) SD _W [k] = W [k] * SD [k]
The representative value SD _W [k] is calculated by. Here, in the above equation (3), k is the window position (window index), and [k] in SD _W [k], W [k] and SD [k] is the value at the window position k, respectively. Indicates that there is.

The signal generation determination unit 125 uses the representative value SD _W calculated by the representative value calculation unit 124 to determine whether or not a biological signal has been generated. For example, if the calculated representative value SD _W [k] exceeds a predetermined threshold value, it may be determined that some biological signal has been generated at the window position k. Further, as a modification mode, which will be described in detail later with reference to FIG. 5, when the time series data of the representative value SD _W shows a predetermined hysteresis of n degrees (n is a natural number), it is determined that the biological signal is generated n times. You can also do it.

Similarly, in FIG. 1, the biological signal discrimination unit 126 determines the type of the generated biological signal in the time interval related to the determination that the biological signal has been generated. For example, it is determined that the generated biological signal is due to the binding motion or the mouth angle raising motion. The processing by the biological signal discriminating unit 126 will be described in detail later when the overall flow is overviewed using the flowchart of FIG.

By the way, the information of the determination result in the signal generation determination unit 125 and the biological signal determination unit 126 is transmitted to, for example, the mobile terminal 2 carried by the user via the signal interface 127, and is used in various applications in the mobile terminal 2. Is also preferable. In this case, for example, the number of biological signal generations per unit time (1 hour, 1 day, etc.) (for example, the number of "smiles" in the case of a myoelectric signal related to raising the angle of the mouth) and the amount of biological activity (for example, "smile"). The amount of muscle contraction activity) related to "" may be displayed as a time-series graph. Further, such information may be recorded as a log, and further, it can be displayed on a display (not shown) provided in the lens portion of the glasses 1 with an EMG sensor.

The signal interface 127 and the mobile terminal 2 are communicated and connected wirelessly or by wire (cable). Of these, the wireless can be, for example, a wireless LAN such as Bluetooth (registered trademark) or Wi-Fi (registered trademark). Further, the wired (cable) may be connected by USB (Universal Serial Bus). The mobile terminal 2 can be a smartphone, a mobile phone, a PDA (Personal Digital Assistant), a tablet computer, or the like, but may be another information processing device such as a personal computer.

Hereinafter, an example in which the biological signal processing according to the present invention is carried out using an input signal including an actual biological signal will be described.

2 and 3 are graphs for explaining an example in which the biological signal processing according to the present invention is carried out. In this embodiment, the glasses 1 with an electromyographic sensor shown in FIG. 1 are used as the biological signal processing device.

In FIG. 2 (A),
(A) "Myoelectric signal / noise: None": Waveform of the input signal with no myoelectric signal or noise,
(B) "Myoelectric signal / noise: Yes": Waveform of myoelectric signal and input signal in which noise is present,
(C) "Myoelectric signal: Yes": Myoelectric signal exists (no noise exists) Input signal waveform, and (d) "Noise: Yes": Noise exists (No myoelectric signal exists) The waveform of the input signal is shown.

For these input signal waveforms, TGAM1 manufactured by Neurosky is used, and a signal sampled at a sampling frequency of 512 Hz is extracted every 128 samples every few minutes, and the extracted unit is used as one point and the point is defined as a line segment. It is the waveform of the connected line graph. In addition, in order to remove the induced noise including the AC component derived from the commercial power source, the notch filter processing is performed in advance in TGAM1. Further, it is known that the above-mentioned "noise" is actually noise generated when the electrodes are displaced due to swelling of the skin or the like.

Of these, according to the input signal waveforms of "myoelectric signal / noise: yes" and "myoelectric signal: yes" in FIG. 2 (A), myoelectric signals caused by muscle activity (such as raising the corner of the mouth and binding) Is found to have a very wide frequency component.

On the other hand, according to the input signal waveform of “Myoelectric signal / noise: None” in FIG. 2 (A), it can be seen that the signal waveform has a small amplitude when there is no expression.

Further, according to the input signal waveform of “Noise: Yes” in FIG. 2 (A), the noise caused by the electrode displacement has a very large amplitude (in some cases, the amplitude is saturated beyond the measurement range). It can be seen that it shows a signal waveform.

Further, in FIG. 3 (A),
(E) "Signal by blinking": The waveform of the input signal when blinking strongly is also shown. According to this input signal waveform, it can be seen that when the eyeglass-type device is used as in the present embodiment and the blinking is strong, artifacts are mixed.

Next, the LPF unit 122b (FIG. 1) performs LPF processing on these input signals shown in FIGS. 2 (A) and 3 (A), and then the acceleration component generation unit 123 (FIG. 1) 2 The results of performing the order difference filter processing are shown in FIGS. 2 (B) and 3 (B).

Of these, in the case of "myoelectric signal / noise: yes" and "myoelectric signal: yes" in FIG. 2B, the waveform of the acceleration component has a form in which a section having a large amplitude is continuously generated. You can see that.

On the other hand, in the case of "myoelectric signal / noise: none" and "noise: yes" in FIG. 2 (B) and "signal due to blinking" in FIG. 3 (B), the waveform of the acceleration component has continuous sections with small amplitude. It can be seen that the shape is formed by.

Here, further, the frequency distribution in the amplitude (acceleration component) of the graphs shown in FIGS. 2 (B) and 3 (B) is shown in FIGS. 2 (C) and 3 (C). In these histograms, the two perpendiculars extending to the positions sandwiching the zero point of the acceleration component indicate the positions of the plus standard deviation and the minus standard deviation, respectively, as seen from the average value position of the same amplitude. There is.

Of these, in the case of "myoelectric signal / noise: yes" and "myoelectric signal: yes" in FIG. 2C, the distance between the two perpendiculars is relatively wide. That is, it can be seen that the standard deviation SD of the frequency distribution becomes larger corresponding to the large periodicity of the amplitude of the acceleration component. On the other hand, in the case of "myoelectric signal / noise: none" and "noise: yes" in FIG. 2C and "signal due to blinking" in FIG. 3C, the distance between the two perpendiculars is relatively narrow. ing. Therefore, the standard deviation SD of the frequency distribution becomes smaller corresponding to the continuity of the sections with small amplitudes for the acceleration component.

In the present invention, it is determined whether or not a "myoelectric signal" other than "noise" is generated by utilizing the property of the amount related to the degree of bias such as the standard deviation SD.

In this regard, conventionally, when detecting an electromyographic signal of, for example, "raising the corner of the mouth" using a glasses-type myoelectric sensor, a sharp artifact with a large amplitude due to "blinking" as shown in FIG. There was a problem that it was erroneously judged as an electromyographic signal of "raising the corner of the mouth". On the other hand, by performing signal processing focusing on the standard deviation SD of the acceleration component as described above, for example, an artifact related to "blinking" can generate a myoelectric signal as a small "noise" of the standard deviation SD. It is possible to take measures such as not including it.

By the way, it is considered that the above-mentioned artifact related to "blinking" is caused by a change in the electrode contact surface pressure due to strong blinking. Similar to such artifacts, artifacts caused by changes in electrode contact surface pressure due to quick movements of the head such as quick nodding, sudden turning, and sudden tilting have actually been detected.

On the other hand, the signal processing according to the present invention focusing on the standard deviation SD of the acceleration component can significantly reduce the error of determining these artifacts as myoelectric signals. That is, it is possible to suppress erroneous detection of bias fluctuations caused by changes in electrode contact pressure and short-time pulsed high bias fluctuations.

Next, the derivation of the weight W used to calculate the representative value SD _W will be described.

In FIG. 2B, in "myoelectric signal / noise: none" in which muscle activity is not active and "noise: yes" in which only noise due to electrode misalignment is observed, the acceleration component is continuous within a predetermined range. The time interval in which the acceleration component stays, in other words, the time interval in which the absolute value of the amplitude of the acceleration component is within the predetermined threshold (len_th interval shown in the figure) is "myoelectric signal / noise: yes" and "muscle". It can be seen that it is considerably longer than the same time section in "Electric signal: Yes".

Therefore, by determining a "weight W" that rapidly decreases as this time interval becomes longer (or at least becomes a monotonically decreasing function for this time interval), and incorporating such a characteristic into the representative value SD _W , the muscle It is possible to more reliably avoid the erroneous determination that the myoelectric signal is generated when there is no electric signal or when there is only noise.

The derivation of the weight W will be specifically described below. Here, first, the amplitude of the acceleration component is scanned from the beginning of the window analysis section to be weighted, and the sample number length len_th in which the amplitude less than the preset threshold value th is continuous is determined. In addition, the number of observation samples obs that defines the noise interval is set in advance. For example, th = 10 and obs = 15 can be set.

Incidentally, when there are a plurality of amplitude continuous sections having a threshold value less than th in the window analysis section, the sample number length len_th may be the total number of samples in those sections. Alternatively, the number of samples in the longest amplitude continuous interval among them can be set to the sample number length len_th.

FIG. 4 is a graph for explaining the weight W used to calculate the representative value SD _{W according} to the present invention.

FIG. 4A shows the exponential weight W as a function of the sample length len_th. This exponential weight W is calculated by the following equation (4) W = exp (1−len_th / obs).
Is stipulated by. Further, in FIG. 4 (A), obs = 15, and if Len_th = 15, W = 1.0. Furthermore, as len_th increases, the exponential weight W decreases sharply and asymptotically approaches zero. In fact, when len_th corresponds to the window analysis interval length (128 samples), the exponential weight W becomes almost zero.

On the other hand, FIG. 4B shows the inverse proportional weight W as a function of the sample number length len_th in the same case of obs = 15. The inverse proportional weight W is calculated by the following equation (5) W = 1 / ((len_th−obs) / a + 1)
Specified by. Here, it is preferable that a takes a value exceeding obs (a> obs) and sets the denominator to a positive value. The inverse proportional weight W in FIG. 4B is the case of a = obs * obs (= 225), and if Len_th = 15, W = 1.0. Also, as len_th increases, the inverse proportional weight W decreases and approaches zero.

Of course, the weight W is not limited to those shown above. A monotonically decreasing function of len_th can be adopted as a weight W, and preferably, a function approaching zero as len_th increases, and more preferably a function asymptotic to zero, various functions are used as weight W. be able to. For example, the weight W may be a linear function of len_th having a negative slope, but the exponential weight W shown in FIG. 4A contributes to a more reliable determination of biological signal generation.

Here, the representative value calculation unit 124 (FIG. 1) weights the calculated standard deviation SD by using the weight W derived as described above, and the above equation (3): SD _W [k] = The representative value SD _W [k] is determined by W [k] * SD [k] (k: window position). Next, the signal generation determination unit 125 (FIG. 1) determines that some biological signal has been generated at the window position k if the calculated representative value SD _W [k] exceeds a predetermined threshold value. ..

Hereinafter, a method of utilizing the hysteresis of the time series data of the representative value SD _W will be described as a preferred modification mode of this signal generation determination.

FIG. 5 is a graph showing an embodiment of signal generation determination processing using the hysteresis of the representative value SD _W. The graph of FIG. 5 is a line graph connecting the time-series data points of the calculated representative value SD _W with a line segment.

The signal generation determination unit 125 (FIG. 1)
(A) The point at which the representative value SD _W (the polygonal line indicating the transition) crosses the threshold Thh line from the lower side (the smaller value) to the upper side (the larger value) is set as the count start point (circle).
(B) The point at which the representative value SD _W (the polygonal line indicating the transition) crosses the line with the threshold value Thr (<Thh) from above (larger value) to downward (smaller value) is the end point (triangle). Mark), as
The count is incremented by 1 for each pair of these count start points and the subsequent count end points.

In the embodiment of FIG. 5, since there are four of these sets, it is determined that the biological signal has been generated four times at the stage when these four sets are determined by the graph (the number of biological signals generated is determined). It counts as 4). Here, by appropriately setting the threshold value (thh) of the start point and the threshold value (thl) of the end point, it is possible to prevent chattering of the signal generation determination result.

Further, when the count end point is not determined even after a predetermined time threshold T _max has elapsed after the signal generation determination unit 125 determines the count start point, the signal generation determination unit 125 performs once from this count start point to that point. After counting, after the lapse of this time threshold value T _{max, the} noise determination period may be set until the representative value SD _W falls below the threshold value Thr. In this case, in the embodiment of FIG. 5, it is finally determined that the number of times the biological signal is generated is five.

[Biomedical signal processing method]
FIG. 6 is a flowchart showing an outline of an embodiment of the biological signal processing method according to the present invention. The flow of the present embodiment shown in FIG. 6 includes a biological signal type determination process by the biological signal determination unit 126 (FIG. 1), and this type determination process will be described in detail separately.

(S101) The acquired input signal is differentially amplified and digitized for buffering.
(S102) A notch filter process for reducing periodic noise related to a commercial power source is performed on the buffered input signal.
(S103) LPF processing is performed on the input signal that has been subjected to the notch filter processing.
(S104) The second-order difference filter processing is performed on the input signal subjected to the LPF processing.
(S105) The input signal that has been subjected to the second-order difference filter processing and has become the acceleration component data of the time series is divided into a predetermined window analysis section.

(S106) The weight W is calculated for each window analysis section.
(S107) The representative value SD _W is calculated for each window analysis section.
(S108) For each window analysis section, it is determined whether or not the representative value SD _W exceeds the predetermined threshold value Th (SD _W > Th).
(S109a) When it is determined in step S108 that the number exceeds (SD _W > Th), it is determined that some biological signal has been generated in the window analysis section, and the process proceeds to step S110 to determine the type.

(S110) In the window analysis section where it is determined that the biological signal has been generated, the biological signal type determination is performed. This determination process will be described in detail later.
(S111) The number of occurrences of the biological signal whose type is determined, that is, the number of occurrences of the corresponding biological phenomenon is counted. This count will also be described later.
(S109b) On the other hand, when it is determined in step S108 that the number does not exceed (SD _W ≤ Th), it is determined that no biological signal is generated in the window analysis section, and this is reflected in the number of times the biological signal is generated. The process proceeds to step S111.

Hereinafter, the biological signal type determination process and the number of occurrence count process in steps S110 and S111, which are carried out by the biological signal determination unit 126 (FIG. 1), will be described.

As an embodiment of this type determination, the biological signal discrimination unit 126 calculates the average power frequency (MPF, mean power frequency) of the input signal in the time interval related to the determination that the biological signal has been generated, and has a high MPF. It is also preferable to determine the type of the generated biological signal based on the above.

Here, the inventor of the present application performs frequency analysis processing by fast Fourier transform (FFT, Fast Fourier Transform) or the like on the input signal acquired by using the glasses 1 with myoelectric sensor, and performs MPF in each window analysis section. As a result of the calculation, it was found that the type of the generated myoelectric signal can be determined by the threshold determination of the MPF value.

Specifically, for example, when the MPF value exceeds a predetermined threshold value, it is determined that the generated biological signal is a myoelectric signal due to a clenching operation, while the MPF value is equal to or less than this predetermined threshold value. , The generated biological signal can be determined to be a myoelectric signal due to the movement of raising the angle of the mouth. It is also possible to generate a hysteresis graph with the vertical axis of FIG. 5 as the MPF value, and similarly count the number of occurrences of the biting motion and the mouth angle raising motion (the myoelectric signal).

Further, in the window analysis section, it is possible to calculate the signal strength, for example, the standard deviation SD'of the amplitude, and use this value for determining the type of the generated signal in the same manner as the MPF value.

By the way, in general, a considerable amount of calculation is required for frequency analysis processing such as FFT, but in the present embodiment, this is performed only in the analysis section where it is determined that some biological signal is generated in steps S108 and S109. Since the frequency analysis is performed, the amount of calculation can be significantly reduced even though the frequency analysis process is performed in the type determination.

Further, as another embodiment of the biological signal type determination, the biological signal discrimination unit 126 includes the standard deviation SD'and the MPF value (MPF) of the input signal in the time interval related to the determination that the biological signal has been generated. A feature amount, for example {SD', MPF}, was calculated, and for this feature amount, the degree of separation, which is the degree of separation from the unit space set by the feature amount of the input signal corresponding to the reference state, was calculated and calculated. It is also preferable to determine the type of the generated biological signal based on the degree of separation.

In this case, specifically, the biological signal discrimination unit 126 combined the state in which the biological signal is generated and the state in which the biological signal is not generated from the degree of separation from the unit space related to the reference state in which the biological signal is not generated. The generation of the biological signal can be determined based on the amount obtained by subtracting the degree of separation from the unit space related to the reference state and the degree of separation from the unit space related to the reference state in which the biological signal is generated.

Here, as the above unit space and the degree of separation,
(A) A value calculated from the unit space in the MT (Mahalanobis Taguchi) method and the Mahalanobis distance.
(B) A value calculated from the unit space in the MTA (Mahalanobis-Taguchi Adjoint) method and the Mahalanobis distance.
A value calculated from (c) the unit space and the characteristic value in the T method, or (d) a value calculated from the unit space and the RT distance in the RT (Recognition Taguchi) method can be adopted. By the way, the inventor of the present application has also found through experiments that such a method for determining the type of biological signal is effective.

Of these, when the MT method is used, for example, when discriminating the myoelectric signal due to the operation of raising the angle of the mouth,
(A) Expressionless state and mouth angle raised state (combined state group)
(B) Expressionless state (c) Design three unit spaces for the raised mouth angle state, and calculate the degree of separation from these unit spaces in the input signal as distance 1, distance 2, and distance 3, respectively. It is known that a more suitable determination result can be obtained by setting (distance for determination) = (distance 2)-(distance 1)-(distance 3).

When such a determination distance exceeds a predetermined threshold value, it can be determined that the generated biological signal is a myoelectric signal due to the mouth angle raising operation. Further, it is also possible to generate a hysteresis graph in which the vertical axis of FIG. 5 is the determination distance, and similarly count the number of occurrences of the mouth angle raising operation (due to the myoelectric signal). Further, for other types of biological signals, for example, myoelectric signals due to the clinging motion, the determination distance can be determined in the same manner as described above, and the type determination and the number of occurrence counts can be performed.

[Other Embodiments of Biological Signal Processing Device]
7 and 8 are schematic views showing another embodiment of the biological signal processing device according to the present invention.

FIG. 7 shows a headphone 1'as a biological signal processing device according to the present invention. The headphone 1'is a wearable device linked to the mobile terminal 2, and can input an input signal that can include a myoelectric signal as a detected biomedical signal, as in the glasses 1 with a myoelectric sensor (FIG. 1). The process is performed in No. 12, the presence / absence of the myoelectric signal generation and the type of the generated myoelectric signal are determined, and the information related to this determination result is transmitted to the mobile terminal 2 via wireless or wired (cable).

Here, the wireless can be, for example, a wireless LAN such as Bluetooth (registered trademark) or Wi-Fi (registered trademark). Further, the wire may be connected to, for example, an analog audio input / output terminal (jack) for headphones / microphones of the mobile terminal 2, or may be connected by USB (Universal Serial Bus). In any case, the audio signal of the content is transmitted from the mobile terminal 2 to the headphone 1'via the wireless or wired system, and the myoelectric message detected by the myoelectric sensor is transmitted from the headphone 1'to the mobile terminal 2. Judgment result information related to the item is transmitted.

In addition, the myoelectric sensor of the headphone 1'also has "detection + (plus) electrode", "reference- (minus) electrode", and "DRL (Driven Right)" as in the case of glasses 1 with myoelectric sensor (FIG. 1). It has three electrodes, "Leg) electrodes". As for the arrangement of these electrodes, as shown in FIG. 7, the reference electrode is in contact with one skin position from around the left (or right) pinna to the vicinity of the cheek, and the detection electrode is in the right (or left) ear. It can be set to touch one skin position near the cheek from the periphery of the auricle. Incidentally, in this case, the biological signals that can be detected include a myoelectric signal caused by a mouth angle raising motion or a clenching motion.

Of course, the arrangement of the electrodes of the myoelectric sensor is not limited to the above-mentioned form. For example, if the headphones 1'do not have an open-air earcup or earpad, the electrodes are placed closer to the cheeks from the surface of the support mechanism that contacts the skin around the ears to attach the headphones to the head. You may.

Further, as a biological signal processing device according to the present invention, as a head-mounted device provided with a similar myoelectric sensor and an electrode thereof, the earphone 1 ″ shown in FIG. 8 can be mentioned. This earphone 1'' also processes an input signal that may include a myoelectric signal as a detected biological signal, determines whether or not a myoelectric signal is generated and the type of the generated myoelectric signal, and uses this determination result as a result. Such information is transmitted to the mobile terminal 2 via wireless or wired (cable). Further, for example, an audio signal of content is transmitted from the mobile terminal 2 to the earphone 1 ″ via the wireless or wired system.

Using the glasses 1 with myoelectric sensor (FIG. 1), headphones 1'(FIG. 7), and earphones 1'' (FIG. 8), which are described above and are worn at positions including the ears, for example, " It is also possible to detect an electromyographic signal related to a mouth angle raising movement corresponding to a facial expression including "smile". By the way, such a myoelectric signal can be used as a user interface when it is a signal due to a user's conscious reaction. On the other hand, if the signal is an unconscious reaction, it can be used as a measurement result of the user's emotion and its transition.

For example, the mobile terminal 2 transmits the sound of the content being played to the headphone 1', and by detecting the myoelectric signal during the voice experience of the user wearing the headphone 1', the user holds the content. It becomes possible to acquire information related to emotions. In addition, the user can determine whether or not the headphones 1'is worn / not worn from the detection status of the myoelectric signal.

By the way, the glasses 1 with myoelectric sensor (Fig. 1), headphones 1'(Fig. 7), and earphones 1'' (Fig. 8), which are worn at the position including the ears, have only myoelectric signals from the muscles in the head. It is also possible to detect biological signals related to body temperature, sweating, pulse wave, pulse, brain wave, etc., which can be detected from a position near the ear. The biological signal processing method of the embodiment as described above can be applied not only to the processing of myoelectric signals but also to the processing of various types of biological signals.

In particular, more effective processing can be performed on noisy AC signals such as myoelectric signals and brain waves. That is, the detection is performed by utilizing the property that the signal detected by the myoelectric sensor using the dry electrode is AC, while the AC signal having a small amplitude is not detected, and the noise due to the deviation of the dry electrode ( Since the artifact) is not detected as an myoelectric signal, the myoelectric signal as an AC signal can be detected more reliably while reducing the amount of calculation.

As described in detail above, the present invention calculates a representative value in acceleration component data instead of performing frequency analysis, which generally requires a large amount of calculation, in determining whether or not a biological signal is generated. And perform the judgment process. As a result, according to the present invention, frequency analysis is not required in determining the generation of biological signals, and robust biological signal processing is realized even for the generation of noise due to electrode misalignment.

Further, by using the more accurate determination processing according to the present invention, as one application example, when a myoelectric signal caused by a muscle related to raising the angle of the mouth such as the zygomaticus major muscle is used as a determination target, " "Smile" can be quantitatively measured, and for example, the fun of laughter electronic content can be quantified from the number of smiles and the amount of activity to raise the mouth angle. In addition, command instructions from the user triggered by the user's mouth angle raising operation, such as camera shutter operation and zooming, and favorite registration of the content being viewed can be executed. Furthermore, it is possible to constantly carry out quantitative measurement of "smile" to quantify whether or not the user is living a healthy life.

Furthermore, the present invention also makes it possible to specify various biological signals other than the above-mentioned myoelectric signals and more reliably determine the occurrence of biological phenomena related to the biological signals. Therefore, these judgment results and the measurement results of the number of occurrences can be utilized for evaluation of various types of contents, user interface of biological phenomena by will, and quantification of physical state and emotional / mental state. is there.

With respect to the various embodiments of the present invention described above, various changes, modifications and omissions in the technical idea and scope of the present invention can be easily made by those skilled in the art. The above explanation is just an example and does not attempt to restrict anything. The present invention is limited only to the scope of claims and their equivalents.

1 Glasses with myoelectric sensor (biological signal processor)
1'Headphones (biological signal processor)
1'' Earphone (biological signal processor)
11 Signal processing box 12 Biometric signal processing unit 121 Signal conversion unit 122 Pre-filter processing unit 122a Notch filter unit 122b LPF unit 122b
123 Acceleration component generation unit 123a Second-order difference filter unit 124 Representative value calculation unit 125 Signal generation determination unit 126 Biological signal discrimination unit 127 Signal interface 13 Positive electrode pad 14 Negative electrode pad 15 Nose pad Electrode unit 2 Mobile terminal

Claims (9)

A biological signal processing device that processes an input signal that may include a myoelectric signal.
Acceleration component generation means that applies second-order difference filter processing to the input signal to generate acceleration component data as the second-order difference of the input signal, and
A monotonous decrease in the value indicating the degree of bias of the data in the predetermined time interval in the acceleration component data and the length of the time interval in which the acceleration component of the acceleration component data in the predetermined time interval remains continuously within the predetermined range. A weight that serves as a function is calculated, and a value indicating the degree of bias of the data is weighted by the weight to calculate a representative value related to the degree of bias of the data, or the predetermined time interval in the acceleration component data. A representative value calculation means in which a value indicating the degree of bias of the data in the above is used as a representative value related to the degree of bias of the data.
A biological signal processing device comprising a signal generation determining means for determining that the myoelectric signal has been generated when the representative value exceeds a predetermined threshold value. Further, a pre-filter processing means for performing a band-stop filter process for reducing noise related to a commercial power supply and a low-pass filter process for extracting a bias fluctuation component with respect to the input signal before generating the acceleration component data. The biometric signal processing apparatus according to claim 1, wherein the biosignal processing apparatus is provided. In the signal generation determination means, the time-series data of the representative value moves upward across the line of the first threshold value from below with the passage of time, and then the line of the second threshold value smaller than the first threshold value. 1 or 2 is characterized in that it is determined that the myoelectric signal has been generated n times when the hysteresis as a general behavior that crosses from above to downward is shown to be n degrees (n is a natural number). The biometric signal processor according to the description. A living body that calculates the average power frequency of the input signal in the time interval related to the determination that the myoelectric signal has been generated, and determines the type of the generated myoelectric signal based on the height of the average power frequency. The biological signal processing apparatus according to any one of claims 1 to 3, further comprising a signal discriminating means. As a feature amount related to the generation of the myoelectric signal, an amount including the standard deviation and the average power frequency of the input signal is defined, and then
A state and muscle in which a certain type of myoelectric signal is generated based on any one of the MT (Mahalanobis Taguchi) method, the MTA (Mahalanobis-Taguchi Adjoint) method, the T method, and the RT (Recognition Taguchi) method. Depending on the feature amount related to the input signal in the state where no electric signal is generated and no artifact as noise, which is a signal other than the certain type of myoelectric signal, is generated . A first unit space is set, a second unit space is set according to the feature amount related to the input signal in a state where no myoelectric signal is generated, and a certain type of myoelectric signal is further set. A third unit space is set according to the feature amount related to the input signal in the generated state, and the third unit space is set.
In the time interval related to the determination that some myoelectric signal has been generated, the feature amount related to the input signal is calculated, and the calculated feature amount is used in the first unit space, the second unit space, and the third unit space. The first degree of separation, the second degree of separation, and the third degree of separation, which are the degrees of separation from each of the unit spaces of, are calculated.
When the amount obtained by subtracting the first degree of separation and the third degree of separation from the second degree of separation exceeds a predetermined threshold value, it is considered that some myoelectric signal is a certain type of myoelectric signal. The biological signal processing apparatus according to any one of claims 1 to 4, further comprising a biological signal determining means for determining. The myoelectric signal is a device attached to the user's head, with the reference electrode in contact with one skin position from around the left (or right) pinna to near the cheek, and the detection electrode on the right (or right). The biometric signal according to any one of claims 1 to 5, wherein the signal is obtained by a device having an electrode configuration such that the area around the auricle of the left) is in contact with one skin position near the cheek. Processing equipment. The biometric signal processing device according to any one of claims 1 to 6, wherein the myoelectric signal is a myoelectric signal caused by a muscle related to raising the angle of the mouth. A program that activates a computer mounted on a device that processes input signals that may include myoelectric signals.
Acceleration component generation means that applies second-order difference filter processing to the input signal to generate acceleration component data as the second-order difference of the input signal, and
A monotonous decrease in the value indicating the degree of bias of the data in the predetermined time interval in the acceleration component data and the length of the time interval in which the acceleration component of the acceleration component data in the predetermined time interval remains continuously within the predetermined range. A weight that serves as a function is calculated, and a value indicating the degree of bias of the data is weighted by the weight to calculate a representative value related to the degree of bias of the data, or the predetermined time interval in the acceleration component data. A representative value calculation means in which a value indicating the degree of bias of the data in the above is used as a representative value related to the degree of bias of the data.
A biological signal processing program characterized in that a computer functions as a signal generation determination means for determining that a myoelectric signal has been generated when the representative value exceeds a predetermined threshold value. It is a biological signal processing method by a computer mounted on a device that processes an input signal that may include a myoelectric signal.
A step of performing a second-order difference filter processing on the input signal to generate acceleration component data as a second-order difference of the input signal, and
A monotonous decrease in the value indicating the degree of bias of the data in the predetermined time interval in the acceleration component data and the length of the time interval in which the acceleration component of the acceleration component data in the predetermined time interval remains continuously within the predetermined range. A weight that serves as a function is calculated, and a value indicating the degree of bias of the data is weighted by the weight to calculate a representative value related to the degree of bias of the data, or the predetermined time interval in the acceleration component data. The step of using a value indicating the degree of bias of the data in the above as a representative value related to the degree of bias of the data, and
A biological signal processing method comprising a step of determining that a myoelectric signal has been generated when the representative value exceeds a predetermined threshold value.

Images