JPH10262942A - Satisfaction feeling measuring system and feedback device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a satisfaction feeling measuring system that measures the satisfaction feeling of worker and feeds back the result to the worker or machine. SOLUTION: A neural network 5 in three-layer structure composed of input, intermediate and output layers is learnt by a BP method with the power spectrum of sampled brain wave diagnosed as the state of satisfaction or no satisfaction based on an objective index as an input signal. This neural network 5 inputs the power spectrum of brain wave of worker and evaluates the state of satisfaction or no satisfaction. The evaluation of this neural network 5 is fed back to the worker 1 as very fine vibrations by a feedback part 7.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention relates to the construction of a human-centered production system that satisfies the functional requirements of being proud and joyful in making things, realizing economic efficiency, and being friendly to people. The present invention relates to a satisfaction measurement system for measuring satisfaction of a subject and feeding back the result to an operator or a machine, and more particularly to a system for measuring human satisfaction using a neural network using an electroencephalogram.

[0002]

2. Description of the Related Art The present inventors are proud of the pleasure and joy of making a new production system of the next generation, have economic efficiency, and have human-centered production with three functional requirements of being gentle to people. The system has been proposed. In order to realize and evaluate such a human-centered production system, it is necessary to evaluate the satisfaction of workers actually working at the work site. However, satisfaction in work includes satisfaction due to pleasure to make and satisfaction due to completed pleasure that reflects difficulties and a sense of achievement, and cannot be measured as an index of simple human pleasure-discomfort. .

On the other hand, physiological indices measured in the engineering field include brain waves in the central nervous system and electrodermal activity, blood pressure, heart rate, respiration, pulse waves and the like in the autonomic nervous system.
Among them, the autonomic nervous system has a problem that the initial value is not stable due to the influence of circadian rhythm and the like, and the central EEG is the most reliable index. In particular, in order to investigate the effects of stimuli such as music and fragrance on human brain waves, the power, appearance rate, fluctuation coefficient, etc. of the α-wave when receiving the stimulus while the eyes are constrained Often required.

On the other hand, electroencephalograms are based on the type of
Since the results vary greatly depending on the state of the eyes open and closed, the measurement results with the eyes closed are the results when the eyes are closed, and this is the normal state when humans open their eyes when opening the eyes. It is not considered applicable.

In connection with this, it has been pointed out that the α wave of the brain wave is related to human comfort and discomfort.
A method for evaluating human emotions based on wave fluctuation characteristics has been proposed (Journal of the Japan Society of Mechanical Engineers 403-406, 1995).
May, JP-A-8-17199). Here, a method of estimating the emotional state using a neural network as the input and output of the α-wave fluctuation slope and wake-up / emotion is used, suggesting the possibility of application to products and space design. I have. However, this method also considers measurement in the closed eye state, and has a very long time resolution of about 5 minutes.

[0006]

However, the satisfaction of the worker is based on the action of a person making an object on the premise of the eye open state, which is closely related to the activity and does not matter whether the eyes are open or closed. However, an index different from human comfort in the space is required. In addition, research on evaluation at several emotional levels has been conducted, such as the evaluation of pleasure and discomfort due to fluctuations described above. No ethical issues have been taken into account or even the way in which the evaluated information is fed back. The temporal resolution of the measurement is also required for feedback.

[0007] On the other hand, there has been a biofeedback method in which information is given to the human visual and auditory senses to forcibly create a certain desirable situation. However, these methods forcibly change human physiological information. May cause physiological breakdown. In addition, it is considered that there is a problem that the information is unilaterally fed back without knowing the meaning of the given information and changes only physiologically.

SUMMARY OF THE INVENTION It is an object of the present invention to provide an index of satisfaction in human work and to provide a technique for measuring satisfaction based on the index and to apply the measured result to the site. In particular, it is an object of the present invention to provide a system that can feed back measured satisfaction information in real time. The present invention also provides a system that obtains physiological information from a worker in an open-eye state without disturbing the work, evaluates the satisfaction of the worker based on this information, and feeds back the evaluation result to the worker without disturbing the work. The purpose is to provide. Further, the present invention provides a system that feeds back to a machine handled by the worker based on the measured worker information (physiological information and operation information) so that the worker can always work in a satisfactory state. Aim.

[0009]

The satisfaction measuring system according to the present invention for achieving the above object comprises a detecting means for detecting an electroencephalogram, a signal processing means for processing a signal from the detecting means to obtain a power spectrum, A neural network that receives a power spectrum and outputs satisfaction information. Preferably, the neural network has a three-layer structure composed of an input layer, a hidden layer, and an output layer,
It is learned by the error back propagation method (hereinafter referred to as BP method).

[0010] This neural network evaluates satisfaction from a plurality of emotional words based on a questionnaire, selects samples determined to be satisfied or unsatisfied based on the evaluation, and evaluates the power spectrum of brain waves in each sample. Learned as an input signal. Particularly, in a neural network having two output layers, the learning is performed so that the output in the satisfactory state is (1.0, 0.0) and the output in the unsatisfactory state is (0.0, 1.0). Satisfaction can be evaluated by determining the relative intensity of the output of one output unit in the total of the output values of the two output units.

The detecting means preferably includes means for detecting and amplifying brain waves by deriving the forehead bipolar and means for A / D converting the amplified signal. Such a detecting means can be easily attached to the worker during the work, and can detect information (brain wave signal) effective for satisfaction measurement without hindering the work.

The satisfaction measuring system of the present invention further comprises means for feeding back the output of the neural network to the worker to be measured or a machine handled by the worker. When the evaluation of satisfaction is fed back to the worker, the feedback unit uses minute vibration as feedback information. This allows the worker to know his / her condition and take measures such as taking a break and changing work conditions without being known to others.

[0013] Such a feedback means is preferably a receiving means for wirelessly receiving the output of the neural network, a switching means triggered by a signal from the receiving means, and an eccentric weight driven by the control of the switching means. And a small motor having the same. By adopting the wireless system, even if the work place is far from the satisfaction measurement system, feedback can be performed without hindering the work.

When the evaluation of the satisfaction is fed back to the machine handled by the worker, the feedback means includes a load, a screen, a sound, and the like on the operation unit of the machine so as to increase the satisfaction of the worker.
Control lighting and the like.

Further, the satisfaction measuring system of the present invention comprises a detecting means for detecting an electroencephalogram, a signal processing means for processing a signal from the detecting means to obtain a power spectrum, and a second means for inputting the power spectrum and outputting satisfaction information. A second neural network that receives a power spectrum as input and outputs information (for example, a degree of noise) related to noise mixed into an electroencephalogram;
Determining means for invalidating the output of the first neural network when the output of the neural network exceeds a predetermined noise value. In this satisfaction measurement system, noise is determined by the second neural network, and only the satisfaction evaluation of the first neural network in a state with less noise is regarded as an effective evaluation. Accurate measurement is possible even in a work environment where noise is expected to be large.

Further, the feedback device of the present invention inputs information such as physiological information of a person or psychological information which is a processing result of the physiological information, and feeds back the person using minute vibrations as feedback information. By utilizing the micro-vibration as feedback information, a person can grasp his / her own state without being known to others. Such a feedback device is preferably a receiving unit that receives the information as a wireless signal from a unit that measures information such as physiological information of a person or psychological information that is a processing result of the physiological information, and a signal from the receiving unit. A switching means having a trigger as a trigger, driven by control of the switching means,
A small motor having an eccentric weight. By using a wireless system, feedback is possible without restricting the behavior.

[0017]

FIG. 1 is a block diagram showing the configuration of an embodiment of a satisfaction measuring system according to the present invention. The satisfaction measuring system mainly includes an electroencephalograph (amplifier) 2 for measuring an electroencephalogram of the worker 1, an A / D converter 3 for digitally converting an analog signal measured by the electroencephalograph 2,
A signal processing unit 4 that performs signal processing such as filter processing and Fourier transform on the digital signal from the A / D conversion unit 3, a neural network 5 that receives a signal from the signal processing unit 4 and outputs a signal indicating satisfaction. , A display unit 6 for displaying the output of the neural network 5, and a feedback unit 7 for feeding back the output of the neural network 5 to the worker 1 and / or a machine handled by the worker 1.

The electroencephalograph 2 is equipped with an amplifier and an A / D converter (A / D conversion unit 3) which are worn by a worker during work and measure brain waves as analog signals. An insulated amplifier whose insulation is insulated by a photocoupler, a transformer, or the like can be used, and it may be a wireless type or a wired type. Although there is a relationship of f _L = 1 / 2πτ between the low-frequency cut-off frequency f _L and the time constant τ of an amplifier generally used as an electroencephalograph, in the case of the system of the present invention, it is intended for a worker who is in motion. The time constant τ is usually set to (0.
3 seconds) and about 0.1 seconds (1.5 Hz) is preferable. The sampling frequency is preferably set to the same frequency as the sampling frequency at the time of measuring the brain wave, for example, 120 Hz in the correlation (described later) between the psychological information and the brain wave information performed as a premise of constructing the neural network.

The electroencephalograph 2 is preferably a one-channel electroencephalograph derived from the forehead bipolar based on the forehead, from the viewpoint of being worn on the worker. The one-channel (single-point measurement) method obtains a smaller amount of information than a multi-channel electroencephalograph, but compensates for its drawback by deriving from the frontal region reflecting emotions. In addition, there is an advantage that mounting is easy because it can be simply wrapped around the head, hardly affected by body movement and noise, and inexpensive.

The A / D converter 3 converts an analog signal measured by the electroencephalograph 2 into a digital signal.
A conversion board can be used. If the above amplifier is not isolated, it must be isolated before entering the board. As the A / D conversion unit 3, in addition to the A / D conversion board, a device (DASport:
Intelligent Instrumentation) can be used. By using this device, measurement with a notebook-type personal computer that is excellent in portability can be performed. Further, as the A / D converter 3, an Ethernet data acquisition device (EDASport: Intelligent Instrumentati) based on the TCP / IP protocol is used.
(manufactured by On Co., Ltd.). In this case, measurement is possible even at a distance of up to 100 m, and measurement via the Internet is possible.

FIG. 2 shows an example of an electroencephalogram measurement system derived from the forehead bipolar. The electroencephalograph 2 has three Ag-AgCl electrodes 21 attached to a band in the form of a button with reference to the center electrode, and the electrode 21 can be arranged on the forehead simply by wrapping the band around the head. it can. The electroencephalograph 2 includes a transmitter 22 having an amplifier and an A / D converter.
After detecting the minute potential by the bipolar extraction with the electrode 21 arranged on the forehead and amplifying it with the amplifier built in the transmitter 22,
For example, the data is digitized at a predetermined sampling frequency, and the data is taken into the computer 9 wirelessly.

The signal processing unit 4 and the neural network 5 are composed of software incorporated in the computer 9, and based on the original signals of the brain waves, the power spectrum, the band power, the appearance rates of α and β waves, noise information, satisfaction information Is calculated, and the result is plotted on the display unit 6 as visual information or sent to the feedback unit 7. As the computer 9, for example, a personal computer level computer incorporating Win3.1 or Win95 can be used.

As shown in FIG. 3, the neural network 5 has a multi-input / multi-output feed-forward type three-layer structure composed of an input layer, an intermediate layer, and an output layer. And the like that have been learned by a known method such as Although the number of units in each layer is not particularly limited, the number of units depends on the frequency resolution because a power spectrum is used as an input signal in the present invention. By increasing the number of units in the intermediate layer, a complicated function can be approximated, but as the number of units increases, the number of local minimum values increases. On the other hand, when learning is established, the capacity of the neural network is not significantly affected by the number of units in the hidden layer. Therefore, the number of units in the intermediate layer may be a minimum number at which learning is possible or a slightly larger number. Specifically, it may be about 20. Although the output layer has two units in FIG. 3, it may be composed of one unit or three or more units.

In the embodiment of the present invention, learning is performed by the BP method using a neural network having 214 units of the input layer, 20 units of the intermediate layer, and 2 units of the output layer to construct a neural network.

Next, an embodiment of the neural network construction according to the present invention will be described.

Prior to the construction of the neural network, modeling of satisfaction at the psychological level was first performed. When engineering satisfaction is measured, it is necessary to obtain correspondence between subjective psychological information based on linguistic reports and physiological information that can be objectively measured. In the present invention, a questionnaire using a multifaceted emotional state scale is performed, considering that psychological information of satisfaction cannot be evaluated by one scale of satisfaction like conventional psychological evaluation. Therefore, the relationship between satisfaction and a multifaceted emotional state scale composed of eight basic emotional states (each emotional state is composed of 10 emotional words) was determined. This is because, unlike animals, when evaluating humans who cannot open their brains for examination, psychological information must be evaluated using linguistic questionnaires.

With such a linguistic model of satisfaction, the user is asked how much satisfaction he or she has, such as the method used in the conventional evaluation of psychological information.
Since the definition of ambiguous abstract measures, rather than evaluation on a single scale, can be evaluated by scores of many intuitively comprehensible emotion words consisting of concrete adjectives, It is now possible to evaluate satisfaction closer to the true value.

On the other hand, in an experiment for obtaining a correspondence between satisfaction obtained as psychological information and information obtained from electroencephalogram which is physiological information, an electroencephalogram was measured using an electroencephalogram measurement system as shown in FIG. Was measured as time series data, and power spectrum analysis was performed. Analysis time 8.5
Power spectrum analysis was performed on 1024 points of data every 3 seconds and 512 points of data every 4.27 seconds using a Hanning window.

The brain waves are δ waves (1 to 4 Hz) and θ
Wave (4-8 Hz), α wave (8-13 Hz), β wave (13-3
0Hz) and γ-waves (30Hz-). Low-frequency bands are susceptible to eye movements and high-frequency bands are susceptible to myoelectric potentials. Was analyzed. Spearman's rank correlation coefficient was determined in order to investigate the relationship between the change in the appearance rate, which is the relative power of the α and β waves, and satisfaction. As a result, 1) the appearance rate of the α band is higher (p <0.05) and the appearance rate of the β band is lower (p <
0.05), 2) Increase / decrease in the appearance rate of the α band has a positive correlation with increase / decrease in satisfaction (statistically significant), and increase / decrease in the appearance rate of the β band has a negative correlation (statistics). Significance).

Based on the results of the questionnaire and the brain wave measurement, learning of the neural network was performed by the BP method (FIG. 4). As input signals to the neural network, 10 samples when satisfied and 10 samples when unsatisfied were selected with reference to the above-mentioned index (emotion word evaluation), and a total of 20 samples were selected.
A power spectrum of 5 to 30 Hz including a wave was used as an input signal. Incidentally, although it can be said that the number of samples of 20 is small for use in learning by the BP method, in the case of the present embodiment, a statistically significant value in which the relationship between the input and the output is known is handled, I think there is no problem with 20. On the contrary, there is an advantage in learning that the local minimum value does not fall.

In general, the input / output function to the neural network must be a derivative. Therefore, in this embodiment, the input / output function is an increasing function, and a sigmoid function (formula (1)) which is a continuous nonlinear function is used as an input / output function. Using.

[0032]

(Equation 1) In the equation, ε is a coefficient indicating the slope of the sigmoid function. If the value is small, the shape becomes close to a linear function, and if the value is large, nonlinearity is emphasized. In this embodiment, ε = 0.5. Note that a non-linear function other than sigmoid can be used as the input / output function.

The input value to the unit is 0 of the sigmoid function.
Although it is necessary to take a value in the range of １1 to 1, the sigmoid function is a saturation function. Then, it was standardized to 0.1 to 0.9. By normalizing an input signal, even if a power spectrum is directly input, identification reflecting the appearance rate, which is relative power, can be performed.

FIG. 5 shows an example of an input signal to the neural network. FIG. 5A shows an input signal when the user is satisfied, and FIG. 5B shows an input signal when the user is not satisfied.

As the corresponding teacher signal, the output when the user is satisfied is (1.0, 0.0), and the output when the user is unsatisfied is (0.0, 1.0). And Originally, the teacher signal should use the score of the psychological information obtained from the questionnaire and give an analog value, but the use of digital values in this way means that the linguistic report It is difficult to reflect on As described above, in the present invention, errors in psychological information are reduced by evaluating satisfaction from multiple points based on the scores of emotion words composed of a total of 80 words. Although the feeling can be evaluated, the satisfaction is regarded as a state in the learning, and a digital value is used as the teacher signal. As described above, learning was performed using satisfaction as a state as a teacher signal, and as described later, the satisfaction can be appropriately evaluated by learning.

Under these conditions, the connection weight was modified using the BP method so that the square error between the output of the neural network and the desired output (teacher signal) was minimized.

In a neural network, generally Nk
Is the number of units in the k-th layer and the output value of the j-th unit u ^k _{j in} the ^k- th layer when the pattern p is presented.

[0038]

(Equation 2) The teacher signal of unit j of the last layer n layer is

[0039]

(Equation 3) Then, the least square error is defined by the equation (2).

[0040]

(Equation 4) The amount of correction is obtained by calculating the gradient of the connection weight of the square error, and the amount of correction between the i-th unit of the K-1 layer and the j-th unit of the k-th layer when a certain pattern p is presented.

[0041]

(Equation 5) Is represented by equation (3).

[0042]

(Equation 6) In this embodiment, η = 0.7. In the BP method, a sequential correction method for performing a correction to present one input pattern with respect to the time of correction of the connection load, and a correction amount of the connection load is accumulated when each input pattern is provided, and all input patterns are presented. There is a collective correction method in which correction is collectively performed later. In this embodiment, a collective correction method for obtaining a coupling load that reduces an error for all patterns is used. Further, in order to prevent the divergence of the learning and to speed up the learning, the connection weight was corrected by introducing an inertia term as shown in Expression (4).

[0043]

(Equation 7) Here, m is the number of times of learning, α is a coefficient called an inertia coefficient, and α = 0.5 in the present invention. The coupling load between the output layer and the intermediate layer, and the coupling load between the intermediate layer and the input layer are expressed by Expressions (5) and (6), respectively.

[0044]

(Equation 8) The learning is terminated when the least square error becomes 1 × 10 ^-4.
In the case of 12-point data, the number was 493 times.

In order to confirm whether learning has been properly performed on the neural network trained in this way,
Evaluation was performed using unlearned data (data not used for learning). One example of the result is shown in FIG. In the figure, (a) shows the evaluation of the unlearned data by the neural network, and (b) and (c) show the appearance rate of the α band and the appearance rate of the β band, respectively. The evaluation of the neural network is the first in the sum of the output values of the two units.
The relative strength of the unit. As can be seen from FIG. 6, the evaluation of the neural network that the satisfaction is high is α
This corresponds to the index of the increase in the appearance rate of the band and the decrease in the appearance rate of the β band, indicating that the learning was appropriate.

The above is an embodiment of the construction of a neural network, and a method other than the BP method can be adopted as long as the neural network has been appropriately trained.

In the satisfaction measuring system of the present invention, satisfaction is measured using the neural network learned in this manner. Preferably, the same electroencephalograph as used in the construction of the neural network is provided to the operator. Attach and measure. The analog signal from the electroencephalograph 2 is converted into a digital signal by the A / D converter 3, and the digital signal is Fourier-transformed by the signal processor 4 to obtain a power spectrum. Further, the appearance rates of the α band and the β band are obtained as necessary. These results can be displayed on the display unit 6. The neural network 5 after learning uses the power spectrum from the signal processing unit 4 as an input signal and obtains outputs corresponding to satisfaction and dissatisfaction from the first and second units. The computer 9 incorporating the neural network 5 determines, as satisfaction, the relative strength of the first unit with respect to the sum of the output values of the two output units.

The satisfaction measuring system according to the present invention is capable of measuring the satisfaction generated by the pleasure of the worker to make things, and furthermore, feeds the measurement result back to the worker or the machine handled by the worker. It is possible to increase the satisfaction of the worker and realize an ideal work state centered on humans.

Hereinafter, as a first mode of the feedback unit 7, the feedback unit 7 for feeding back the evaluation of satisfaction by the neural network 5 to the worker will be described.

As the feedback section 7, for example, a voice or a display on a monitor can be adopted. However, since the worker performs work by visual and audible information, these may hinder the work. However, there is a possibility that there is a problem that the management is not known to others and leads to alienation of humanity. Therefore, in consideration of the fact that the work of the worker is not affected, the information is intuitively transmitted to the worker, and there is no problem in terms of ethics, etc. Is preferred. Thus, the worker can grasp his / her own state without interruption.

As the result obtained by the satisfaction measuring system, there are a state in which the user is satisfied and a state in which the user is not satisfied. Either of them may be fed back. Both can be fed back with different information. By knowing his / her condition, the worker can take a break when he is not satisfied, continue working when he is satisfied, or play music that is good for himself. Satisfaction feedback tells you when you are happy and can be used for training.

FIG. 7 shows an embodiment of a system for feeding back the result obtained by the satisfaction measuring system to the worker by touch. The feedback unit 7 includes a transmission unit 71 that wirelessly transmits a result obtained by the satisfaction measurement system (computer 9) as a predetermined electric signal, and a transmission unit 71.
And a switching circuit 73 that operates by a signal from the receiving unit 72 and a vibration generator 74. As shown in FIG. 8, the vibration generator 74 includes a small motor 75 and a weight 76 attached to an eccentric shaft 75a of the motor.
6 oscillates and generates very small vibrations. Receiving unit 72,
The switching circuit 73 and the vibration generator 74 are configured as an integrated feedback device, and can be put in a case and attached to a worker's body. Therefore, even at a position distant from the computer 9, the measurement result can be fed back by a wireless signal from the transmission unit 71 attached to the computer 9.

The feedback section 7 operates, for example, as follows. When the evaluation in the satisfaction measurement system is in an unsatisfied state, for example, if a value below a certain value continues or a state in which the rate of change is negative continues, a signal is output from the transmission unit 71. When the receiving section 72 receives such a signal, the computer 9 sends an on / off signal to the switching circuit 73. As a result, the switching circuit 73 operates to turn on the motor 75. As a result, the worker feels slight movement, does not affect the work, knows his / her current state without anyone's knowledge, can take a break, and can change work conditions.

When the operator wants to know the state of satisfaction, the signal transmitted from the transmitting section 71 may be changed by changing the software.

Next, feedback to a machine will be described as a second mode of the feedback section. FIG. 9 is a diagram schematically showing a feedback system 80 for a machine. This system is a system for feeding information from an operator to a machine system as well as the satisfaction measurement system described above. This feedback system is based on a state evaluation unit 81 that measures and evaluates various information from an operator, and based on information from the state evaluation unit 81, understands the state of the person and evaluates the best value for the evaluation item. It comprises a state comprehension unit 82 that sends a signal for controlling each part of the machine (for example, a handle, a light, a motor, etc.) to each part, and a controller 83 for each part.

The state evaluation section 81 includes physiological information such as brain waves, heartbeats, pulse waves, blood pressures, skin potentials, and skin temperatures, and behavior information such as facial expressions, body movements, and hand movements. Information) and the physical quantity such as the amount of operation on the mechanical system and the pressure change applied to the operating part (together with the operation information)
Is input, the state of the worker is evaluated, and this is sent to the state understanding unit 82 as a predetermined signal. For example, it is possible to determine the degree of fatigue of the worker from the heart rate, blood pressure, and the like, and to determine the mental state including the degree of satisfaction from the brain waves. The state evaluation unit 81 may be simple software when the determination content is simple, but a neural network is appropriate when complicated evaluation such as satisfaction measurement is involved. The state evaluation unit 81 corresponds to the electroencephalograph 2, the signal processing unit 4, and the neural network 5 in the satisfaction measurement system of FIG. That is, when the information obtained from the worker is an electroencephalogram, the state evaluation section 81 evaluates the satisfaction by measuring the electroencephalogram, and sends the result to the state understanding section 82.

The state understanding unit 82 controls the controller 83 to operate each part of the mechanical system so that the evaluation item finally takes the best value. In this case, it is not always necessary to know the relationship between the amount of operation or feedback to the sense and the state to be evaluated, and the state understanding unit 82 may search for the amount of feedback so that the state is improved by the feedback. . However, it is effective to know a certain degree of correspondence between the measured information and the evaluation items.

By employing such an exploratory feedback system, it is difficult for an operator who finds a unique relationship between a state identified by individual differences, a state of the day, and the like, and operation information or environmental information. Can be targeted.

What is controlled by the signal from the state understanding unit 82 includes, for example, the load on the operation handle, the driving speed, the display on the monitor, the illumination, the sound and the like. When the target machine is a personal computer, the display time of characters and the like by input from a keyboard, the response speed of a mouse and a pointing device, the screen display, the sound, and the like are included. These can be controlled by adjusting them on software. For example, when the worker is satisfied, it is predicted that the work can be performed efficiently, so that the display is performed quickly. If you are unsatisfied, change the screen slightly, such as the size of the arrow, the design of the icon, and the color, to give a variety of work. Or the music that the worker likes. This shows that the machine cares about its condition, and the familiarity makes the work fun.

When the target machine is an NC machine tool, in addition to controlling the screen display and sound of the monitor as in the case of the personal computer described above, the weight of the jog handle for manual operation and the 1 The amount of movement of the memory can also be controlled. For example, as for the weight of the jog handle, the current value of the spindle motor is taken out, input to a magnet powder brake for controlling the weight of the jog handle, and a cutting torque substantially proportional to the cutting torque is given to the jog handle. As a result, a sense of unity between the operator and the machine comes out, the friendliness of the machine springs up, and the work can be enjoyed. It is also possible to make the handle heavier or lighter in response to satisfaction. This gives diversity in work. Further, as the screen display, the processing phenomenon may be displayed in real time in accordance with the result of the satisfaction measurement. This is made possible by capturing an image of the cutting point with a CCD camera while synchronizing with the rotation of the workpiece.

When the target machine is a factory line, control of the line speed and the supply amount to the line can be mentioned. For example, according to the degree of satisfaction, the line speed can be changed so as not to affect the production, or the line can be controlled to collect parts for a person who is satisfied.

As described above, the satisfaction measurement system including feedback not only to the worker but also to the machine can be applied to a machine system that takes an action adapted to the satisfaction of human beings, and is ideal for humans. A system realizing a machine interface can be obtained.

Next, as a second embodiment of the present invention, a satisfaction measuring system further incorporating a noise discrimination neural network will be described.

The satisfaction measurement system of the present invention is intended for workers who may be at work sites where the environment is not very good. In this case, noise generated from body movements, machines, and the surrounding environment is taken into consideration. There is a need to. Further, even in a place where there is no noise, noise may be generated by the electrode device. In the second embodiment, the presence or absence of these noises is identified by the second neural network, and the reliability of the evaluation by the satisfaction measurement neural network is determined based on the result.

FIG. 10 is a block diagram showing the configuration of a second embodiment of the satisfaction measuring system according to the present invention. In FIG. 10, the same components as those of the satisfaction measuring system of FIG. 1 are denoted by the same reference numerals, and description thereof will be omitted. This satisfaction measurement system differs from the satisfaction measurement system of FIG. 1 in that it includes a neural network 8 for judging noise of an electroencephalogram signal in parallel with a neural network 5 for evaluating satisfaction. The neural network 8 receives the power spectrum from the signal processing unit 4 as input and determines whether there is noise equal to or greater than a threshold.

As the noise discrimination neural network 8, a feedforward type three-layered neural network similar to the satisfaction measurement neural network 5 and learned by the BP method or the like can be used. Like the neural network 5, the neural network 8 can also be constructed as software in the same computer.

As an example, a neural network having 172 units of the input layer, 20 units of the intermediate layer, and 2 units of the output layer was constructed. As the teacher signal, 10 samples each with and without noise were used, a total of 20 samples.

In general, an electroencephalogram is a signal having a meaning up to about 30 Hz, and a signal having a frequency of 40 Hz or more is greatly affected by a potential generated by body motion or the like. Therefore, when constructing the neural network 8, the power spectrum of the signal in the band of 40 to 60 Hz is used as an input signal, and when the output signal has no noise (1.0, 0.0), when the noise has occurred ( 0.0, 1.0). The coefficients η, α, and ε for learning are 0.7, 0.5,
The learning was terminated when the least square error became 1 × 10 ⁻⁴ . The number of times of learning was 6441 times and 7910 times for two types of data (1024 data points and 512 data points) having different power spectrum analysis times, respectively.

In the satisfaction measuring system incorporating such a noise discrimination neural network, an analog signal from the electroencephalograph 2 is converted into an A / D converter 3 as in the embodiment of FIG.
To convert the digital signal into a digital signal, and Fourier transform the digital signal in the signal processing unit 4 to obtain a power spectrum. Further, the appearance rates of the α band and the β band are obtained as necessary. These results can be displayed on a display unit (not shown). Each neural network 5, 8
Using the power spectrum as an input signal, the former evaluates satisfaction as in the embodiment of FIG. If it is determined that the user is unsatisfied, the computer 9 sends a signal to the receiving unit via the transmitting unit of the feedback unit 7, whereby the vibration generator of the feedback unit 7 is activated, and the operator is caused by the minute vibration. Give your status feedback to. Also in this case, it is possible to feed back a state of satisfaction.

On the other hand, the neural network 8 evaluates the presence or absence of noise. In this case, the computer 9 (determination unit 10) obtains the degree of noise as a ratio of the second unit to the sum of the output values of the two output units of the neural network 8, and if the value is equal to or larger than a predetermined threshold, , It is determined that the reliability of the result of the neural network 5 is low, and the signal from the transmitting unit is not transmitted regardless of the result. This can be similarly applied to the case where the result evaluated as satisfactory in the neural network 5 is fed back.

Therefore, only the evaluation of the satisfaction determined by the neural network 5 is measured in a state with little noise, so that extremely accurate measurement can be performed.

In the above embodiment, the case where the neural network is constructed as software of a computer has been described. However, the neural network can be constructed as a single chip.

In the above embodiment, the case where the satisfaction measurement system is developed using the data during the work and the satisfaction of the worker is mainly measured has been described. If considered as a stimulus, the present invention can be applied to general situations as well as simple tasks. For example, the present invention can be applied to the care of a bedridden sick or disabled person in a hospital.

[0074]

As is apparent from the above description, the present invention provides a system for objectively evaluating satisfaction in work. The results of this satisfaction measurement system can be used as feedback information to workers and / or machines, thereby enabling human-centered production system design.

The satisfaction measuring system according to the present invention incorporates a second neural network for noise identification in addition to the satisfaction evaluation neural network.
A highly reliable system can be provided.

The satisfaction measuring system of the present invention is provided with an electroencephalograph which can be easily worn, so that the worker can voluntarily work without interrupting the worker's work without the help of an electroencephalogram measurement specialist. , And satisfaction can be measured.

Further, the feedback device of the present invention utilizes minute vibration as feedback information,
Since there is no ethical problem such as the use of the result of satisfaction measurement or the like for management purposes and the user can intuitively grasp his / her own state, a person can spontaneously create an ideal working state.

[Brief description of the drawings]

FIG. 1 is a configuration diagram schematically showing an embodiment of a satisfaction measurement system according to the present invention.

FIG. 2 is a diagram showing an embodiment of a detecting means in the satisfaction measuring system of the present invention.

FIG. 3 is a diagram showing an embodiment of a structure of a neural network of the satisfaction measuring system according to the present invention.

FIG. 4 is a flowchart showing a learning process of the neural network.

5A and 5B are diagrams showing an example of an input signal to the neural network, wherein FIG. 5A shows power spectrum data in a satisfactory state, and FIG. 5B shows power spectrum data in an unsatisfactory state.

6A and 6B are diagrams showing evaluation results of the neural network, wherein FIG. 6A shows the output of the neural network, FIG. 6B shows the α band appearance rate, and FIG. 6C shows the β band appearance rate.

FIG. 7 is a configuration diagram schematically showing a feedback system to a person according to the present invention;

FIG. 8 is a diagram showing a main part of a feedback unit.

FIG. 9 is a configuration diagram schematically showing a feedback system for a machine according to the present invention.

FIG. 10 is a configuration diagram schematically showing another embodiment of a satisfaction measurement system according to the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Worker 2 ... Amplifier (electroencephalograph) 3 ... A / D conversion part 4 ... Signal processing part (signal processing means) 5. ····· Neural network (first neural network) 7 ······ Feedback unit 8 ····· Neural network (first neural network) 9 ···· Computers 22 and 71 ... Transmission means 23, 72 ... Reception means 73 ... Switching means 75 ... Motor 76 ... Eccentric weight

────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] July 30, 1997

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0017

[Correction method] Change

[Correction contents]

[0017]

FIG. 1 is a block diagram showing the configuration of an embodiment of a satisfaction measuring system according to the present invention. The satisfaction measurement system mainly includes an electroencephalograph ( amplifier) 2 for measuring an electroencephalogram of the worker 1, an A / D converter 3 for converting an analog signal measured by the electroencephalograph 2 into a digital signal,
A signal processing unit 4 that performs signal processing such as filtering and Fourier transform on the digital signal from the / D conversion unit 3, a neural network 5 that receives a signal from the signal processing unit 4 and outputs a signal indicating satisfaction. Neural network 5
And a feedback unit 7 that feeds back the output of the neural network 5 to the worker 1 and / or a machine handled by the worker 1.

[Procedure amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0040

[Correction method] Change

[Correction contents]

[0040]

(Equation 4) Correction amount is obtained by calculating the slope of the binding load of square error of the k -1 layer i-th time presented a pattern p unit and between the k layer j-th unit

[Procedure amendment 3]

[Document name to be amended] Statement

[Correction target item name] 0055

[Correction method] Change

[Correction contents]

Next, feedback to a machine will be described as a second mode of the feedback section. Figure 9 is a diagram showing an outline of a feedback system 80 of the machine, the eyes this system feeds <br/> back-to-mechanical system information from the worker generally not only satisfaction measurement system described above system It is. This feedback system is based on a state evaluation unit 81 that measures and evaluates various information from an operator, and based on information from the state evaluation unit 81, understands the state of the person and evaluates the best value for the evaluation item. A state understanding unit 82 that sends a signal to control each part of the machine (eg, steering wheel, lighting, motor, etc.) to be taken.
And a controller 83 for each part.

[Procedure amendment 4]

[Document name to be amended] Statement

[Correction target item name] 0061

[Correction method] Change

[Correction contents]

When the target machine is a factory line, control of the line speed and the supply amount to the line can be mentioned. For example, according to the degree of satisfaction, the line speed can be changed so as not to affect the production, or the line can be controlled to collect parts for a person who is satisfied. The target machines are the above-mentioned computers.
Data, factory lines, etc., and humans such as robots.
Anything that can control it can be applied
Wear.

[Procedure amendment 5]

[Document name to be amended] Statement

[Correction target item name] 0073

[Correction method] Change

[Correction contents]

In the above embodiment, the case where the satisfaction measurement system is developed using the data during the work and the satisfaction of the worker is mainly measured has been described. The present invention can be applied not only to simple tasks but also to general situations. For example, the present invention can be applied to the care of a bedridden sick or disabled person in a hospital. In addition, from home appliances to automobiles and aircraft
The range of application can be expanded by the design development of

Claims (13)

[Claims] 1. A system comprising: detecting means for detecting an electroencephalogram; signal processing means for processing a signal from the detecting means to obtain a power spectrum; and a neural network for inputting the power spectrum and outputting satisfaction information. Satisfaction measurement system. 2. A detecting means for detecting an electroencephalogram, a signal processing means for processing a signal from the detecting means to obtain a power spectrum, a neural network which receives the power spectrum and outputs satisfaction information, and an output of the neural network. A satisfaction measuring system comprising a means for feeding back to a measured object. 3. The neural network includes an input layer,
3. The satisfaction measuring system according to claim 1, having a three-layer structure including an intermediate layer and an output layer. 4. The neural network has an output layer composed of two units, and outputs in a satisfactory state are (1.
0, 0.0), and the output in the unsatisfied state is (0.0, 0.0).
3. The satisfaction measuring system according to claim 1, wherein learning is performed by an error back propagation method so as to be 1.0). 5. The apparatus according to claim 1, wherein said detecting means includes means for detecting and amplifying an electroencephalogram by deriving a forehead bipolar, and means for A / D converting the amplified signal.
The satisfaction measurement system described. 6. The apparatus according to claim 1, wherein said detecting means includes a transmitting means for wirelessly transmitting the detected brain wave signal.
Or the satisfaction measurement system according to 2. 7. The satisfaction measuring system according to claim 2, wherein said feedback means uses minute vibrations as feedback information. 8. The feedback means is driven by control of the receiving means for receiving the output of the neural network as a radio signal, switching means triggered by a signal from the receiving means, and control of the switching means.
The satisfaction measuring system according to claim 7, further comprising a motor having an eccentric weight. 9. A detecting means for detecting an electroencephalogram of an operator working on a machine, a signal processing means for processing a signal from the detecting means to obtain a power spectrum, and a neural network for inputting the power spectrum and outputting satisfaction information. And a means for feeding back the output of the neural network to the machine. 10. The satisfaction measurement according to claim 9, wherein said machine is an NC machine tool, and further comprising means for controlling a jog handle for manual operation based on satisfaction information from said feedback means. system. 11. A detecting means for detecting an electroencephalogram, a signal processing means for processing a signal from the detecting means to obtain a power spectrum, a first neural network which inputs the power spectrum and outputs satisfaction information, and the power spectrum. And a second neural network that outputs information on noise mixed into the brain wave as an input and disables the output of the first neural network when the output of the second neural network exceeds a predetermined noise value. A satisfaction measuring system, comprising: 12. An apparatus for inputting information such as physiological information of a person or psychological information as a result of processing the physiological information and feeding back the information to the person, wherein a minute vibration is used as feedback information. . 13. A device for inputting information such as physiological information of a person or psychological information as a result of processing the physiological information and feeding back the information to the person, wherein an output from a means for outputting the information is received as a radio signal. A feedback device, comprising: a receiving unit that performs the control, a switching unit that uses a signal from the receiving unit as a trigger, and a motor that is driven by the control of the switching unit and that has an eccentric weight.