JP3203535B2 - Exercise evaluation method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion evaluation method for evaluating the motion of an object using jerk obtained by differentiating acceleration.

[0002]

2. Description of the Related Art Physical quantities representing the motion of an object include position,
Speed and acceleration can be mentioned. There is a sensor for detecting each physical quantity. By the way, humans detect the position and the speed visually, and the sensible detection as the bodily sensation is the acceleration and the higher-order physical quantities. For example, when the user is riding on a moving body such as an automobile, a railroad vehicle, an aircraft, or an elevator, he or she is conscious of the state change during acceleration or deceleration.

Development of acceleration sensor for automobile (Nissan Technical Report No.
According to literatures such as No. 23, Sho 62-12), it is known that humans react more sensitively to jerks of higher orders than acceleration. This fact indicates that speed is deeply involved in the sensation of pleasure and discomfort felt by humans. Therefore, in order to evaluate a vehicle involving movement such as an automobile, a railway vehicle, an aircraft, and an elevator, not only this acceleration but also information on jerk obtained by differentiating the acceleration is detected.
It is necessary to evaluate using that information. In other words, conversely, humans limit acceleration to linear motion,
Since it is not possible to detect physical quantities higher than jerk speed,
It can be said that if the acceleration and jerk information can be detected, the amount of information for evaluating the vehicle is sufficient.

[0004] Japanese Patent Laid-Open No.
In Japanese Patent Publication No. 3 (1993), a jerk sensor capable of simultaneously detecting acceleration and jerk and a movement control / motion evaluation method using the jerk sensor are proposed. In this exercise evaluation method, the acceleration and jerk detected at a plurality of detection points are evaluated by a value obtained by multiplying the acceleration and jerk by an arbitrary weighting constant.

[0005]

In the above-mentioned prior art, for example, an object which is steadily vibrating is effective in that the strength of the vibration component is evaluated. There is no guideline for an evaluation method for a movement in which the periodicity in the change of the temporal acceleration and jerk during the movement is sparse. Further, when the target of the motion evaluation changes, the setting of the weighting constant and the algorithm of the evaluation function also need to be changed for each motion evaluation target, which is complicated. On the other hand, assuming that a person feels pleasant and uncomfortable when riding on a car, a railroad car, an aircraft, and an elevator, as a result of evaluation by algorithms such as sensuality and sensibility using acceleration and jerk as inputs, A result such as pleasant or unpleasant is output. Human evaluation has a personality, and if a human changes during a series of exercise evaluations, the subsequent evaluations will be inconsistent, unlike the previous evaluations. In such a case, the exercise evaluation may be restarted from the beginning. Also, even for the same person, the evaluation may change due to a change in physical condition. Therefore, if an algorithm that simulates human sensuality and sensibility can be formed in performing the motion evaluation, the motion evaluation that excludes the personality as well as the human evaluator such as a test driver, together with the jerk detection means Should be possible. Further, according to the above-described exercise evaluation method, an abnormal exercise can be detected from the evaluation content in the same manner as a human evaluator.

It is an object of the present invention to provide a motion evaluation method for evaluating the motion of various objects by jerk.

An object of the present invention is to provide a motion evaluation method for evaluating the motion of an object in the same way as a human sense.

An object of the present invention is to provide a motion evaluation method for evaluating the motion of an object that changes irregularly.

An object of the present invention is to provide a motion evaluation method for evaluating the motion of an object that changes with periodicity.

An object of the present invention is to provide an abnormal motion detecting method for detecting an abnormal motion of an object.

[0011]

[0012]

SUMMARY OF THE INVENTION It is an object of the present invention to detect a jerk at least by differentiating the acceleration of a moving object for each unit time, delay the jerk detected for each unit time, and form a hierarchical structure in order. This is achieved by inputting the jerk pattern to the input layer of the neural network as a jerk pattern and outputting the motion evaluation result from the output layer via the hidden layer.

An object of the present invention is to detect an jerk obtained by differentiating at least the acceleration of a moving object for each unit time, analyze the jerk detected for each unit time by frequency, and apply a frequency to an input layer of the hierarchical neural network. This is achieved by inputting as a spectrum pattern and outputting the motion evaluation result from the output layer via the intermediate layer.

[0014]

[0015]

The object of the present invention is to provide an ideal output, which is an evaluation based on human sensitivity to jerks during various motions of an object, for a connection load between an input layer and a hidden layer and a hidden layer and an output layer of the neural network. By detecting jerks of various movements of the object, inputting the detected jerks to the neural network and changing the output from the output layer so as to reduce the difference between them, and learning the neural network in advance. Achieved.

The weight of the connection between the input layer and the intermediate layer and between the intermediate layer and the output layer of the neural network is calculated based on the ideal output which is evaluated by human sensitivity to the jerk of the motion of the object;
It is desirable that the neural network be learned at any time while changing the difference between the output of the neural network and the output of the neural network when evaluating the motion of the object.

If the motion of the object cannot be evaluated, it is desirable to generate an abnormal motion signal as an unlearned motion.

[0019]

[0020]

[0021]

According to the above arrangement, the jerk of the moving object is detected and the motion is evaluated by the neural network, so that various algorithms for evaluating the motion of the object are not required.

The jerk of an object that changes irregularly is detected at an arbitrary time interval, and the detected jerks are delayed and input to the neural network in order, so that the jerks are arranged at regular intervals in time series. Jerk speed patterning is possible.

A frequency spectrum pattern can be obtained by detecting the jerk of an object whose motion changes periodically and analyzing the frequency.

By pre-learning a neural network using human senses as ideal outputs (teacher signals),
An algorithm that simulates human sensation can be formed, and the motion of an object can be evaluated in the same way as human sensation.

By learning the neural network as needed at the time of evaluating the motion of the object, the motion of a new object can be evaluated.

When the motion of the object cannot be evaluated, an abnormal motion signal is generated as an unlearned motion, thereby making it possible to detect an abnormal motion that was not assumed at the time of design.

By detecting the acceleration as well as the jerk of the moving object, it is possible to perform a motion evaluation that cannot be performed only by jerk detection.

By detecting the jerk with the jerk sensor, the combination of the acceleration sensor and the electric circuit for differentiating the detected value can also detect a frequency region in which the electric circuit does not function due to a phase delay.

[0029]

Embodiments of the present invention will be described below with reference to the drawings.

First, the basic configuration of this embodiment will be described.

FIG. 1 is an explanatory diagram for explaining the exercise evaluation method according to the embodiment of the present invention in comparison with the conventional sensory evaluation.

Conventionally, when performing a sensory evaluation of a vehicle such as an elevator or a vehicle, a tester has sensually or empirically evaluated information that the tester actually got on the vehicle and experienced. The amount of information that can be experienced at this time is only the acceleration and jerk applied to the tester. Therefore, if this acceleration and jerk are detected simultaneously, it is necessary and sufficient as physical information for evaluation.If a means (emulator) that imitates the process of linking the evaluation from this physical quantity can be realized, it will be possible to realize human sensitivity evaluation. A similar evaluation should be possible.

In this embodiment, a jerk sensor for detecting acceleration and jerk is used, and a neural network is used as an emulator. A neural network connects multiple processing units corresponding to nerve cells (neurons) in a network like a neural network of the brain, and exchanges signals with each other, moving many processing units in parallel and This is a mechanism that performs information processing in the entire network, and is an engineering model of the information processing mechanism that is performed in the brains of living organisms such as humans. In addition, the neural network spontaneously changes the strength of the connection between neurons, that is, the characteristics of the network, when the sample data is repeatedly processed, and can gradually output the desired output data from the input sample data. . That is, a desired algorithm can be obtained while learning is repeated even if the clear algorithm is unknown at first.

The neural network learning method for acquiring the algorithm in the present embodiment uses supervised learning, and the input test pattern is acceleration and jerk information detected simultaneously when a human sensory evaluation is performed.
A human sensory evaluation is given as the teacher signal. As the learning is repeated, human sensuality and experience come to be reflected in the neural network. Therefore, a motion evaluation method can be configured by using such a learned neural network and jerk sensors. Further, by changing the exercise to be learned, an evaluation for an arbitrary exercise can be performed.

Next, the configuration of the apparatus of this embodiment will be described.

FIG. 2 is an explanatory diagram showing the configuration of the apparatus according to the embodiment of the present invention.

As shown in the figure, the apparatus is a moving object 0
, The jerk sensor 1 for detecting the acceleration and jerk generated by the movement of the object 0 as needed, and sampling the acceleration output and jerk output of the jerk sensor 1 at arbitrary equal time intervals. Sampler 2, delay element group 3 for delaying sampled acceleration output and jerk output
And a neural network 4 for inputting an acceleration and jerk pattern from the delay element group 3 and outputting a motion evaluation pattern from an output layer, and furthermore, each label indicating a motion evaluation index for the output of the neural network 4 is stored. And an evaluation label storage unit 5.

Next, the exercise evaluation method of the present embodiment will be described.

FIG. 3 is an explanatory diagram for explaining the functions of the sampler and the delay element group in FIG.

Now, it is assumed that the object 0 makes an arbitrary motion and the jerk sensor 1 outputs a time history of acceleration and jerk as shown in FIG. The sampler 2 samples the acceleration and jerk output of the jerk sensor 1 every unit time from the start of exercise. The sampled acceleration and jerk data are sequentially set to the input layer of the neural network 4 by the delay element group 3 as an input acceleration and jerk pattern. In this way, the time series acceleration,
The jerk information can be patterned.

FIG. 4 is an explanatory diagram for explaining a general neuron model.

FIG. 5 is an explanatory diagram for explaining an input conversion function called a general sigmoid function.

The neural network 4 is a network in which neuron models 400 whose concept is shown in FIG. The j-th neuron model 400 is
The value obtained by multiplying the input signal Xi from the i-th neuron by the connection weight Wij corresponding to the strength of the connection with the neuron is (Wi
j · Xi) is added to all neurons expressed by Equation 1 and a value obtained by adding a bias value θj is held as an internal potential Xj.

[0044]

(Equation 1)

[0045]

(Equation 2)

Then, based on the input conversion function of FIG.
Output one output signal Yj. The output signal Yj output from the j-th neuron model 400 is
The input Xj of the k-th connected neuron model. Further, feedback may be given using the output signal as its own input.

FIG. 6 is an explanatory diagram for explaining the mode of the neural network according to the embodiment of the present invention.

As shown in the figure, the neural network 4 has two modes. One is a processing mode for outputting a motion evaluation pattern in accordance with the input acceleration and jerk patterns, and the other is a learning mode for optimizing the neural network 4 itself by learning in accordance with the following algorithm. is there.

FIG. 7 is an explanatory diagram for explaining neural network learning according to the embodiment of the present invention. Learning of the neural network is, specifically, optimization of a connection weight between neuron models. This embodiment shows a supervised learning method. First, acceleration and jerk information generated during the motion of the object 0 are used as input patterns as sample patterns, and a plurality of sets of ideal evaluation output data corresponding to the sample acceleration and jerk patterns are prepared. The evaluation output may be a result of an evaluation performed by a person actually boarding or the like. When the object 0 performs the motion of the sample jerk pattern described above, the evaluation output is determined using information other than jerk information. May be done. The sample acceleration and jerk pattern are input to the neural network 4 to obtain an evaluation output. Here, if the evaluation output data and the ideal evaluation output data of the sample acceleration and jerk pattern coincide, all the connection weights on the path transmitting the signal in the neural network 4 at that time are obtained. Increase the value. Conversely, if they do not match, the values of all the connection weights on the path transmitting the signal in the neural network 4 at that time are reduced. By repeating this using various sample acceleration and jerk data, the neural network 4 gradually becomes
A desired output can be obtained for any input.
Finally, a desired motion evaluation output can be obtained for an arbitrary acceleration and jerk pattern generated by the motion of the object 0.

When the neural network 4 has been trained as described above, the learning mode for learning is not required, and the motion can be evaluated only in the processing mode for processing the input data. Therefore, when the evaluation target and exercise are limited, the neural network 4 may use a device that has been learned and is configured only in the processing mode. Also, when changing the evaluation target and exercise, activate the learning mode at any time even after the end of learning,
If a new sample acceleration, jerk pattern, and ideal evaluation output data are learned, a network can be formed that can cope with a new motion (acceleration, jerk pattern) as needed.

FIG. 8 is an explanatory diagram for explaining a method of evaluating the riding comfort when the elevator is lifted according to the embodiment of the present invention.

The jerk sensor 11 is attached to the car 10 of the elevator, and can detect vertical acceleration and jerk.

Now, the sampler 12 starts sampling simultaneously with the start of the ascent. The sampled acceleration and jerk information are input to the input layer 141 via the delay element group 13, and are subjected to the conversion processes of Expressions 1 and 2 to obtain the intermediate layer 142.
Output from the output layer 43 via Reference numeral 15 denotes an evaluation label storage unit, which stores each label indicating a motion evaluation index for each neuron model in the output layer 143.

FIG. 9 is an explanatory diagram for explaining an example of evaluating ride comfort when the elevator is raised according to the embodiment of the present invention.

As shown in this figure, four levels of exercise evaluation labels of 'good', 'satisfactory', 'unsatisfactory', and 'impermissible' are given. The evaluation of the motion is determined by the output of the output layer 143 for each label. The size of the fill in the square in the figure reflects the output value of each labeled neuron model in the output layer 143, where 1.0 is filled and 0.0 is blank. In case 1, an output of 1.0 is obtained for a neuron model with a label of “good”, and is close to 0.0 for neurons with other labels. Therefore, the motion of the elevator that provides an output value as in Case 1 can be evaluated as “good” motion. In case 2, an output of 0.5 was obtained for the neuron model labeled “good” and “satisfied”
0.5 was also obtained for the neuron model labeled with, indicating that the movement was such that the evaluation was separated and the movement was worth between the two evaluations. In case 3, an output of 0.6 is obtained for a neuron model with a label of “unsatisfied”, 0.6 is output for a neuron model with a label of “impossible”, and a neuron model with a label of “satisfied” is obtained. On the other hand, 0.5 was obtained, and as a result, it can be understood that the exercise was evaluated as “dissatisfied”. In this way, unlike the conventional digital evaluation method, it is possible to perform an evaluation closer to a human sense, such as performing a fuzzy evaluation. Also, by inputting these signals to an elevator motion controller, the ride comfort of the elevator can be improved.

FIG. 10 is an explanatory diagram for explaining a method of recognizing a vehicle motion according to the embodiment of the present invention.

The jerk sensors 211 and 212 are connected to the vehicle 2
0, respectively, in the front-rear direction (x-axis),
The lateral acceleration (y-axis) and jerk can be detected. Now, the sampler 22 starts sampling simultaneously with the start of traveling. The sampled acceleration and jerk information are stored in delay element groups 231, 232, 233, and 234.
Via the input layer 241 of the neural network 24
Is output to the output layer 243 via the intermediate layer 242 through the conversion process of the equations (1) and (2). Reference numeral 25 denotes an evaluation label storage unit that stores labels for recognizing a motion state of each neuron model in the output layer 243. In this embodiment, 'straight', 'turn' (cor
nering) ',' accelerate ',' decelerat '
e) ',' spinning ',' Tyre lo
ck) 'are labeled for recognizing six types of exercise states. The recognition of the movement is determined by the output of the output layer 243 for each label.

FIG. 11 is an explanatory diagram for explaining an example of recognizing a vehicle motion according to the embodiment of the present invention.

As in FIG. 9, the size of the fill in the square in the figure reflects the output value of each labeled neuron model in the output layer 243, where 1.0 is a fill and 0.0 is a blank. is there. Now, in case 1, an output of 1.0 is obtained for the neuron model labeled 'straight-forward', and 0.0 for other labeled neuron models.
It is. Accordingly, the motion of the vehicle 20 that can obtain the output value as in Case 1 can be recognized as a “straight-forward” motion. In case 2, an output of 0.5 is obtained for the neuron model labeled 'straight-forward', and 0.5 is obtained for the neuron model labeled 'turning'. It can be seen that the movement is a cornering with a large turning radius (close to a straight line). In case 3, an output of 0.5 is obtained for the neuron model labeled 'swirl', 0.6 for the neuron model labeled 'acceleration', and 0.6 for the neuron model labeled 'deceleration'. In this case, the vehicle motion is not in the state where the acceleration and jerk are input to the neuron model, and a bias is accidentally input to the neuron model, and the connection between neurons is caused by the bias. Is in a strong state, there is a large output,
When equal outputs of labels of contradictory physical quantities are obtained, it can be understood that it is appropriate to recognize that there is no acceleration and no deceleration, that is, constant acceleration and constant jerk motion with constant jerk motion. In this way, unlike the conventional digital evaluation method, it is possible to perform an evaluation closer to a human sense, such as performing a fuzzy evaluation.

FIG. 12 is an explanatory diagram showing the acceleration and jerk of a general vehicle when spin occurs. FIG. 13 is an explanatory diagram showing acceleration and jerk when a tire lock occurs in a general vehicle.

When the behavior of the vehicle 20 changes, the centrifugal force in the lateral direction of the vehicle during a turn and the decelerating force in the longitudinal direction of the vehicle during a deceleration decrease sharply. By learning such characteristic acceleration and jerk patterns, it is possible to easily detect and recognize changes in the behavior of the vehicle. Athletic performance can be improved.

Next, an evaluation method for a constantly vibrating object will be described.

FIG. 14 is an explanatory view showing a device configuration of another embodiment of the present invention.

The motion evaluation device shown in FIG. 3 includes a jerk sensor 31 fixed to the object 30 and detecting the acceleration and jerk generated as the object 30 moves, and the acceleration output of the jerk sensor 31. A sampler 32 that samples jerk output at arbitrary equal time intervals, a frequency analyzer 33 that performs frequency analysis on the output from the sampler 32, and inputs an acceleration and jerk pattern and outputs a motion evaluation pattern from an output layer. It comprises a neural network 34 and an evaluation label storage means 35 in which each label representing an evaluation index of the movement with respect to the output of the neural network 34 is stored.

FIG. 15 is an explanatory diagram for explaining the functions of the sampler and the frequency analyzer of FIG.

Assume that the object 30 is now arbitrarily moving as shown in this figure, and the acceleration / jerk time history output from the jerk sensor 31 is shown in FIG. The sampler 32 samples the acceleration and jerk output of the jerk sensor every unit time from the start of exercise. The sampled acceleration and jerk data are subjected to frequency analysis such as Fourier transform by the frequency analyzer 33, and are converted into a power spectrum as shown in the figure. The power spectrum for each frequency is input to the input layer of the neural network 34 as an acceleration and jerk spectrum pattern. In this way, the time series acceleration,
The jerk information can be converted into a spectrum pattern. In the neural network 34, the power spectrum of acceleration and jerk information generated during the movement of the object 30 is used as an input pattern as a sample pattern, and a plurality of sets of ideal evaluation output data corresponding to the sample acceleration and jerk spectrum pattern are provided. By preparing and repeating the learning, the neural network 34 gradually obtains a desired output with respect to an arbitrary input. Finally, as in the previously described embodiment, the object 30
A desired motion evaluation output can be obtained for an arbitrary acceleration and jerk pattern generated by the motion.

Although all of the examples described above are cases where evaluation is possible, cases where evaluation is impossible are also conceivable. For example, as in Case 4 shown in FIG. 9, the highest “good” and the lowest “impermissible” are output simultaneously. These cases show the case where the motion of the object to be evaluated has not been learned. When applied to human evaluation, this is considered to be close to the sensation of riding for the first time on a vehicle that has a clearly different movement pattern from those who have experienced riding in the past. Even in such a case,
Using the learning mode of the neural network,
By learning the motion of a new object without changing the algorithm, it is possible to evaluate the motion of the new object in addition to the conventional evaluation target.

In addition, abnormal movement can be detected by utilizing the evaluation inability in reverse.

FIG. 16 is an explanatory view showing the configuration of an abnormal motion detecting apparatus according to another embodiment of the present invention.

The abnormal movements include, for example, abnormal vibrations caused by a child jumping and jumping in a car, and a phenomenon that a brake is not effective due to a failure. The abnormal motion detection device is a fixed jerk sensor 41 that is fixed to the object 40 and detects the acceleration and jerk generated as a result of the movement of the object 40 as needed. The acceleration output and jerk output of the jerk sensor 41 are arbitrary. , A delay element group 43 for delaying the output of the sampler 42, and a neural network for inputting acceleration and jerk patterns output from the delay element group 43 and outputting a motion evaluation pattern from an output layer. A network 44; an evaluation pattern storage means 46 for storing an evaluation pattern used as an ideal output during learning; a neural network 4
Each label representing the evaluation index of the movement with respect to the output of No. 4 is stored, and the inability to evaluate is detected from the output of the neural network 44 and the evaluation pattern used as the ideal output during learning stored in the evaluation pattern storage means 46,
And an evaluation failure detection device 45 that outputs an abnormal movement signal. In the present embodiment, the neural network 44 is given in advance all motion modes that occur during normal operation and the evaluation of the motion, and the learning is completed. That is, during normal operation, the evaluation is output as learned. On the other hand, when an abnormal motion is input to the neural network 44, evaluation becomes impossible. This state is detected from the output layer 343 by the non-evaluation detection device 45, and an abnormal movement signal is generated, whereby the abnormal movement can be judged by itself. Further, it is possible to detect a change in exercise performance due to a change over time or the like.

In this embodiment, a jerk sensor for detecting jerk and acceleration of an object is used, and a process of linking sensory evaluation with a physical sensory quantity obtained when a person rides on an object is realized by a neural network. A method for realizing the same evaluation as a sensory evaluation by a human and a method for detecting an abnormal motion of an object from these evaluations have also been described. In the present embodiment, the jerk sensor is used to detect the acceleration and jerk of the object.However, the present invention is not limited to this. May be used. Also, position, speed,
Further functions may be achieved by taking in image information and the like and inputting it to a neural network.

Further, the evaluation output by the exercise evaluation method of this embodiment may be input to the exercise controller to improve the exercise performance. Further, the exercise control controller may be configured by a neural network common to the exercise evaluation device, and may learn to be optimal based on the evaluation by the exercise evaluation method.

As described above, according to the present embodiment, the motion of the vehicle can be evaluated in the same manner as the human sense by learning the neural network in advance using the evaluation based on the human sense. It is also possible to perform an evaluation in an extreme condition that cannot be performed by a human, for example, a test flight in which G is large and severe, and an evaluation with high reproducibility and consistency can be obtained by eliminating personality.

[0074]

According to the present invention, the jerk of a moving object is detected and the motion is evaluated by a neural network, so that the motion of various objects can be efficiently evaluated without preparing each algorithm. .

By learning the neural network in advance using the evaluation based on the human sensation as an ideal output, an algorithm simulating the human sensation can be formed, and the motion of the object can be evaluated in the same manner as the human sensation.

By detecting the jerk of a moving object, delaying the jerk pattern, and inputting the pattern to a neural network, the movement of the object that changes irregularly can be evaluated efficiently.

By detecting the jerk of a moving object, performing frequency analysis and converting it into a frequency spectrum pattern and inputting it to a neural network, the motion of an object whose movement changes periodically can be efficiently evaluated. be able to.

If it is impossible to evaluate the motion of the object, an abnormal motion signal is generated as an unlearned motion, thereby making it possible to detect an abnormal motion that was not assumed at the time of design.

By detecting the acceleration as well as the jerk of the moving object, it is possible to evaluate the movement which cannot be achieved by the jerk detection alone.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram for explaining the exercise evaluation method according to the embodiment of the present invention in comparison with a conventional sensory evaluation.

FIG. 2 is an explanatory diagram showing an apparatus configuration of an embodiment of the present invention.

FIG. 3 is an explanatory diagram illustrating functions of a sampler and a delay element group in FIG. 2;

FIG. 4 is an explanatory diagram illustrating a general neuron model.

FIG. 5 is an explanatory diagram illustrating an input conversion function called a general sigmoid function.

FIG. 6 is an explanatory diagram illustrating modes of a neural network according to an embodiment of the present invention.

FIG. 7 is an explanatory diagram for explaining neural network learning according to the embodiment of this invention.

FIG. 8 is an explanatory diagram illustrating a method of evaluating ride comfort when the elevator is raised according to the embodiment of the present invention.

FIG. 9 is an explanatory diagram illustrating an example of evaluating ride comfort when the elevator is raised according to the embodiment of the present invention.

FIG. 10 is an explanatory diagram illustrating a method of recognizing a vehicle motion according to an embodiment of the present invention.

FIG. 11 is an explanatory diagram illustrating an example of recognizing a vehicle motion according to the embodiment of the present invention.

FIG. 12 is an explanatory diagram showing acceleration and jerk of a general vehicle when spin occurs.

FIG. 13 is an explanatory diagram showing acceleration and jerk when a tire lock occurs in a general vehicle.

FIG. 14 is an explanatory diagram showing a device configuration of another embodiment of the present invention.

FIG. 15 is an explanatory diagram illustrating functions of the sampler and the frequency analyzer of FIG.

FIG. 16 is an explanatory diagram showing a configuration of an abnormal motion detection device according to another embodiment of the present invention.

[Explanation of symbols]

 0 Object 1 jerk sensor 2 sampler 3 delay element group 4 neural network 5 evaluation label storage means 10 elevator 11 jerk sensor 12 sampler 13 delay element group 14 neural network 15 evaluation label storage means 20 vehicle 24 neural network 25 evaluation label Storage means 30 object 31 jerk sensor 32 sampler 33 frequency analyzer 34 neural network 35 evaluation label storage means 40 object 41 jerk sensor 42 sampler 43 delay element group 44 neural network 45 evaluation failure detection device 46 evaluation pattern storage means 141 Input layer 142 middle layer 143 output layer 211 jerk sensor 212 jerk sensor 221 sampler 222 sampler 223 sampler 224 sampler 231 delay Element group 232 delay element group 233 delay element group 234 delay element group 241 input layer 242 intermediate layer 243 outputs layer 400 neuron model 441 input layer 442 intermediate layer 443 outputs layer

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int. Cl. ⁷ , DB name) G01P 15/00-15/13 G06F 15/18 520 G06F 15/18 550

Claims (5)

(57) [Claims] 1. A jerk obtained by differentiating at least the acceleration of a moving object is detected for each unit time, and the jerk detected for each unit time is delayed and sequentially added to an input layer of a hierarchical neural network. A motion evaluation method characterized by inputting as a speed pattern and outputting a motion evaluation result from an output layer via an intermediate layer. 2. A jerk obtained by differentiating at least the acceleration of a moving object is detected for each unit time, and the jerk detected for each unit time is frequency-analyzed to a frequency spectrum pattern to an input layer of a hierarchical neural network. And outputting a motion evaluation result from an output layer via an intermediate layer. 3. An ideal output, which is an evaluation based on human sensitivity to jerks during various motions of an object, and a connection between an input layer and an intermediate layer and an intermediate layer and an output layer of the neural network. Detecting jerks of various motions, inputting the detected jerks to the neural network and changing the output from an output layer so as to reduce a difference between the jerks and the neural network, and learning the neural network in advance. The exercise evaluation method according to claim 1 or 2, which performs the exercise evaluation. 4. An ideal output, which is an evaluation based on human sensitivity to the jerk of the motion of an object, and a connection weight between an input layer and an intermediate layer, and an intermediate layer and an output layer of the neural network. claim 1 also characterized by any time learning the neural network is changed so that the difference becomes smaller between the output of the neural network
Is the exercise evaluation method according to 2. 5. When the motion of the object cannot be evaluated,
Motion estimation method according to claim 1 or 2, characterized in that cause abnormal motion signal as unlearned movement.