CN-116322508-B - Behavior state estimation device, behavior state estimation method, behavior state learning device, and behavior state learning method

Abstract

A behavior state estimation device (10) is provided with a sampling unit (11), a statistic calculation unit (12), a behavior state model storage unit (13), and an estimation calculation unit (14). A sampling unit (11) samples the displacement measurement signal for a predetermined time period, and generates displacement measurement data. A statistic calculation unit (12) calculates a statistic of the displacement measurement data. A behavior state model storage unit (13) stores a behavior state model obtained by modeling a statistic and a desired muscle load state by associating them. The estimation calculation unit (14) uses the statistics as an input vector and uses the behavior state model to estimate the load state.

Inventors

• MATSUMOTO YOSHIHIKO
• NAITO ATSUSHI
• KAWAHARA NAOKI
• GAO WANTAI

Assignees

• 株式会社村田制作所

Dates

Publication Date

20260512

Application Date

20210721

Priority Date

20200730

Claims (20)

1. 1. A behavior state estimation device is provided with: A first sampling unit that samples a displacement measurement signal, which is a signal obtained by converting displacement of the skin surface of a subject due to physiological tremor and deformation caused by muscles into a voltage, within a predetermined period of time, and generates displacement measurement data; a first statistics calculation unit that calculates a first statistics of the displacement measurement data; A behavior state model storage unit for storing a behavior state model obtained by modeling the first statistics and the load state of the muscle by correlating them, and A presumption calculation unit presuming the load state by using the action state model and using the first statistics as input vectors, The action state model sets the importance of the statistic for each muscle to be estimated of the load state, The importance is an importance for determining a weight when calculating a final estimation result from a plurality of the load states.
2. 2. The behavior state estimation device according to claim 1, wherein, The statistic calculation unit calculates the first statistic from a signal intensity distribution obtained by arranging the displacement measurement data in order of increasing signal intensity.
3. 3. The behavior state estimation device according to claim 2, wherein, The statistic calculation unit divides the signal intensity distribution into a plurality of signal intensity blocks, and uses the statistic calculated for each of the signal intensity blocks as the first statistic.
4. 4. The behavior state estimation device according to claim 1, wherein, The statistic calculation unit blocks the displacement measurement data at predetermined times arranged in time series, and uses the statistic calculated for each block as the first statistic.
5. 5. The behavior state estimation device according to claim 1, wherein, The behavior state estimation device further comprises: A second sampling unit for sampling an operation measurement signal generated by detecting an operation of the subject by an operation detection sensor and output the signal, within a predetermined period of time, to generate operation measurement data, and A second statistic calculation unit that calculates a second statistic for the motion measurement data, The behavior state model storage unit stores a behavior state model obtained by modeling by correlating the first statistics, the second statistics, and the load states of the muscles, The estimation operation unit uses the first statistic and the second statistic as input vectors, and estimates the load state using the behavior state model.
6. 6. The behavior state estimation device according to claim 5, wherein, The second statistic calculation unit divides the motion measurement data into a plurality of blocks, and uses the statistic calculated for each of the blocks as the second statistic.
7. 7. The behavior state estimation device according to claim 5 or 6, wherein, The action state model sets the importance of the first statistic and the importance of the second statistic for each muscle to be estimated of the load state.
8. 8. The behavior state estimation device according to claim 7, wherein, The importance of the first statistic and the importance of the second statistic are set by a common importance.
9. 9. The behavior state estimation device according to claim 7, wherein, The importance of the first statistic and the importance of the second statistic are set separately.
10. 10. The behavior state estimation device according to claim 5 or 6, wherein, The motion measurement signal is a measurement signal of at least one of acceleration and angular velocity.
11. 11. The behavior state estimation device according to claim 1, wherein, The displacement measurement signal is a signal obtained by voltage-converting a displacement amount including an influence caused by physiological tremor.
12. 12. The behavior state estimation device according to claim 1, wherein, As the input vector, biological information is also included.
13. 13. A behavior state estimation method includes: A first sampling process of sampling a displacement measurement signal, which is a signal obtained by converting displacement of the skin surface of the subject due to physiological tremor and deformation caused by muscles into a voltage, within a predetermined time period, to generate displacement measurement data; a first statistical calculation process of calculating a first statistical quantity of the displacement measurement data, and A presumption calculation process of presuming the load state by using an action state model obtained by modeling by correlating the first statistic and the load state of the muscle and using the first statistic as an input vector, The action state model sets the importance of the statistic for each muscle to be estimated of the load state, The importance is an importance for determining a weight when calculating a final estimation result from a plurality of the load states.
14. 14. The method for motion state estimation according to claim 13, wherein, The action state estimation method further comprises the following steps: a second sampling process of sampling an operation measurement signal generated by detecting an operation of the subject by an operation detection sensor and output the operation measurement signal, for a predetermined period of time, to generate operation measurement data, and A second statistic calculation process of calculating a second statistic for the motion measurement data, In the estimation operation process, the load state is estimated using an action state model obtained by modeling a first statistic and a second statistic by correlating the load state of the muscle, and using the first statistic and the second statistic as input vectors.
15. 15. A behavior state learning device is provided with: A first sampling unit that samples a displacement measurement signal, which is a signal obtained by converting displacement of the skin surface of a subject due to physiological tremor and deformation caused by muscles into a voltage, within a predetermined period of time, and generates displacement measurement data; a first statistics calculation unit that calculates a first statistics of the displacement measurement data; A third sampling unit that samples a muscle activity measurement signal composed of a muscle potential obtained by detecting a muscle activity of the subject for a predetermined period of time, and generates muscle activity measurement data; a modeling reference value calculation unit for calculating a modeling reference value of the behavior state model from the muscle activity measurement data, and A learning operation unit that generates the behavior state model by performing learning using the first statistics and the modeling reference value, The action state model sets the importance of the statistic for each muscle to be estimated as a load state, The importance is an importance for determining a weight when calculating a final estimation result from a plurality of the load states.
16. 16. The behavior state learning device according to claim 15, wherein, The learning operation unit generates the behavior state model by synchronizing the first statistics with the modeling reference value.
17. 17. The behavior state learning device according to claim 15 or 16, wherein, The behavior state learning device further comprises: A second sampling unit for sampling an operation measurement signal generated by detecting an operation of the subject by an operation detection sensor and output the signal, within a predetermined period of time, to generate operation measurement data, and A second statistic calculation unit that calculates a second statistic for the motion measurement data, The learning operation unit generates the behavior state model using the first statistic, the second statistic, and the modeling reference value.
18. 18. The behavior state learning device according to claim 17, wherein, The learning operation unit synchronizes the first statistic, the second statistic, and the modeling reference value to generate the behavior state model.
19. 19. A behavior state learning method includes: A first sampling process of sampling a displacement measurement signal, which is a signal obtained by converting displacement of the skin surface of the subject due to physiological tremor and deformation caused by muscles into a voltage, within a predetermined time period, to generate displacement measurement data; A first statistical calculation process of calculating a first statistical quantity of the displacement measurement data; a third sampling process of sampling a muscle activity measurement signal composed of a muscle potential obtained by detecting a muscle activity of the subject for a predetermined time to generate muscle activity measurement data; a modeling reference value calculation process of calculating a modeling reference value of the behavior state model from the muscle activity measurement data, and A learning operation process of performing learning using the first statistics and the modeling reference value to generate the behavior state model, The action state model sets the importance of the statistic for each muscle to be estimated as a load state, The importance is an importance for determining a weight when calculating a final estimation result from a plurality of the load states.
20. 20. The method for learning behavior state according to claim 19, wherein, The action state learning method further comprises the following steps: a second sampling process of sampling an operation measurement signal generated by detecting an operation of the subject by an operation detection sensor and output the operation measurement signal, for a predetermined period of time, to generate operation measurement data, and A second statistic calculation process of calculating a second statistic for the motion measurement data, And generating a behavior state model using the first statistic, the second statistic, and the modeling reference value in the learning operation process.

Description

Behavior state estimation device, behavior state estimation method, behavior state learning device, and behavior state learning method Technical Field The present invention relates to a technique for estimating an action state including a load state of a muscle from a detection result of tremors, and a technique for generating an action state model (action state learning technique) used for the estimation technique. Background Patent document 1 describes an action state estimating device and the like. The behavior state estimation device described in patent document 1 converts a measurement signal of a displacement detection sensor into a frequency component. The behavior state estimating device described in patent document 1 estimates a behavior state from components of a predetermined frequency band. Patent document 1 Japanese patent application laid-open No. 2011-182824 However, in the conventional device and method shown in patent document 1, it is necessary to convert the measurement signal of the sensor into a frequency component. Therefore, the processing load for generating the signal for estimation increases. Disclosure of Invention Accordingly, an object of the present invention is to provide a behavior state estimation technique that achieves a desired estimation accuracy and suppresses a processing load. A behavior state estimation device is provided with a first sampling unit, a first statistic calculation unit, a behavior state model storage unit, and an estimation calculation unit. The first sampling unit samples a displacement measurement signal of the subject for a predetermined time period, and generates displacement measurement data. The first statistic calculation unit calculates a first statistic of the displacement measurement data. The behavior state model storage unit stores a behavior state model obtained by modeling the first aggregate and the desired load state of the muscle by associating them. The estimation calculation unit uses the first statistics as an input vector and uses the behavior state model to estimate the load state. In this configuration, the load state is estimated using measurement data that is not converted into a frequency component. According to the present invention, it is possible to achieve a desired estimation accuracy and to suppress a processing load. Drawings Fig. 1 is a functional block diagram of an action state estimating device according to a first embodiment of the present invention. Fig. 2 is a table showing an example of setting the importance of the first embodiment. Fig. 3 is a flowchart showing the main processing of the behavior state estimation method according to the first embodiment of the present invention. Fig. 4 is a functional block diagram of the behavior state learning device according to the first embodiment of the present invention. Fig. 5 is a flowchart showing main processing of the behavior state learning method according to the first embodiment of the present invention. Fig. 6 (a) shows a time-varying example of signal intensity, fig. 6 (B) shows a signal intensity distribution, and fig. 6 (C) shows intensity block data. Fig. 7 is a table showing an example of setting the importance of the second embodiment. Fig. 8a is a diagram showing a time-varying example of signal intensity, fig. 8B is a diagram showing intensity block data (average value), and fig. 8C is a diagram showing intensity block data (integrated value). Fig. 9 is a functional block diagram of an action state estimating device according to a third embodiment of the present invention. Fig. 10 (a) and 10 (B) are tables showing examples of setting of importance in the third embodiment. Fig. 11 is a flowchart showing main processing of the behavior state estimation method according to the third embodiment of the present invention. Fig. 12 is a functional block diagram of an action state learning device according to a third embodiment of the present invention. Fig. 13 is a flowchart showing main processing of the behavior state learning method according to the third embodiment of the present invention. Fig. 14 (a) and 14 (B) are tables showing examples of setting of importance in the fourth embodiment. Fig. 15 is a flowchart showing the main processing of the behavior state learning method according to the fifth embodiment of the present invention. Fig. 16 is a diagram showing a concept of synchronization. Detailed Description (First embodiment) The behavior state estimation technique and the behavior state model generation technique according to the first embodiment of the present invention will be described with reference to the drawings. (Structure and processing of action State presumption device) Fig. 1 is a functional block diagram of an action state estimating device according to a first embodiment of the present invention. As shown in fig. 1, the behavior state estimation device 10 includes a time-series sampling unit 11, a statistic calculation unit 12, an est