Information processing device, program, and information processing method

Abstract

The signal processing device (100) includes a signal acquisition unit (131) that acquires a time series signal indicating a plurality of waveforms, a signal shaping unit (132) that identifies one waveform from the time series signal, and a frequency analysis unit (133) that calculates a power parameter that is a parameter indicating the power of the one waveform.

Description

[Technical field]

[0001] The present disclosure relates to an information processing device, a program, and an information processing method. [Background technology]

[0002] In recent years, technology has been developed for analyzing signals that have components that repeatedly fluctuate up and down, such as pulse waves, each of which has a different waveform.

[0003] The pulse wave analysis device described in Patent Document 1 detects the minimum and maximum values ​​of the pulse wave, divides the waveform value into units equivalent to one beat, and plays back the waveform value at high speed multiple times in beat units to obtain the spectrum of the pulse wave. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 7-148126 Summary of the Invention [Problem to be solved by the invention]

[0005] Pulse waves are influenced by hemodynamic biomechanics such as cardiac contractility, aortic valve closure, and systemic vascular resistance. It is known that the pulse wave period, pulse wave power, and pulse wave shape of a pulse wave signal not only indicate whether the pulse is normal or abnormal, such as arrhythmia or atrial fibrillation, but also indicate emotional states as a result of autonomic nervous function. For example, autonomic nervous function can be evaluated by analyzing the pulse wave power.

[0006] However, while the conventional technology calculates the spectrum of a pulse wave by processing a pulse wave signal, it does not disclose or suggest the extraction and use of pulse wave power. Therefore, the conventional technology has a problem in that it cannot take into account the fluctuation components in one waveform due to the biomechanics of hemodynamics, and therefore cannot perform accurate analysis.

[0007] Therefore, an object of one or more aspects of the present disclosure is to make it possible to calculate power parameters for each wave from a time-series signal. [Means for solving the problem]

[0008] An information processing device according to an aspect of the present disclosure includes: a signal acquisition unit that acquires a time-series signal indicating a plurality of waveforms; and an identification unit that identifies one waveform from the time-series signal. Calculated by taking the square root of the sum of the squares of the frequency spectrum, a power parameter which is a parameter indicating the power of the one waveform, a period parameter which is a parameter indicating the period of the one waveform, and The frequency spectrum is normalized by the square root of the sum of squares. and a calculation unit that calculates the waveform shape parameters.

[0009] A program according to an aspect of the present disclosure includes a signal acquisition unit that acquires a time-series signal indicating a plurality of waveforms, a determination unit that determines one waveform from the time-series signal, and Calculated by taking the square root of the sum of the squares of the frequency spectrum, a power parameter which is a parameter indicating the power of the one waveform, a period parameter which is a parameter indicating the period of the one waveform, and The frequency spectrum is normalized by the square root of the sum of squares. The present invention is characterized in that the signal processing unit functions as a calculation unit for calculating waveform shape parameters.

[0010] An information processing method according to an aspect of the present disclosure includes acquiring a time-series signal indicating a plurality of waveforms, identifying one waveform from the time-series signal, Calculated by taking the square root of the sum of the squares of the frequency spectrum, a power parameter which is a parameter indicating the power of the one waveform, a period parameter which is a parameter indicating the period of the one waveform, and The frequency spectrum is normalized by the square root of the sum of squares. The method is characterized by calculating waveform shape parameters. Effect of the Invention

[0011] According to one or more aspects of the present disclosure, power parameters can be calculated for each wave from a time series signal. [Brief description of the drawings]

[0012] [Figure 1] 1 is a block diagram illustrating a schematic configuration of a signal processing device according to a first embodiment. [Diagram 2] FIG. 2 is a schematic diagram showing an example of a waveform represented by a time-series signal. [Diagram 3] FIG. 1 is a block diagram showing a schematic configuration of a computer. [Figure 4] 13 is a flowchart showing a frequency analysis process. [Diagram 5] FIG. 1 is a diagram showing a schematic diagram of a waveform shape and power decomposition. [Figure 6] FIG. 11 is a block diagram illustrating a schematic configuration of an emotion determination system according to second and third embodiments. [Figure 7] FIG. 11 is a block diagram illustrating a schematic configuration of a learning device 200 according to a second embodiment. [Figure 8] FIG. 11 is a block diagram showing a schematic configuration of an inference device according to second and third embodiments. [Figure 9] FIG. 11 is a block diagram illustrating a schematic configuration of a learning device according to a third embodiment. [Figure 10] FIG. 11 is a block diagram illustrating a schematic configuration of a signal reconstruction device according to a fourth embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Embodiment 1 FIG. 1 is a block diagram illustrating a schematic configuration of a signal processing device 100 which is an information processing device according to the first embodiment. The signal processing device 100 includes a detection unit 110, a connection I / F (InterFace) unit 120, a control unit 130, and a parameter storage unit 140. Although both the detection unit 110 and the connection I / F unit 120 are provided in FIG. 1, the signal processing device 100 only needs to include at least one of them.

[0014] The detection unit 110 detects a physical quantity related to the object, and provides the control unit 130 with a time-series signal that corresponds to the physical quantity and has a component that repeatedly fluctuates up and down, such as a pulse wave.

[0015] The connection I / F unit 120 is a functional unit that can connect to an external storage device (not shown) and can transfer data to and from the external storage device. The external storage device is an optical disk or a flash memory storage device. Here, it is assumed that the above-mentioned time-series signal is stored in the external storage device. The connection I / F unit 120 reads out the time-series signal stored in the external storage device and provides the time-series signal to the control unit 130.

[0016] The control unit 130 controls the processing in the signal processing device 100 . The control unit 130 includes a signal acquisition unit 131, a signal shaping unit 132, a frequency analysis unit 133, and a parameter storage unit 134.

[0017] The signal acquiring unit 131 acquires a time-series signal from the detecting unit 110, or from an external storage device via the connecting I / F unit 120. The acquired time-series signal is provided to the signal shaping unit 132. The time-series signal is a signal that indicates a plurality of waveforms.

[0018] The signal shaping unit 132 is a determination unit that determines a waveform from the time-series signal from the signal acquiring unit 131 and sequentially generates a waveform signal that is a signal corresponding to the determined waveform. The generated waveform signal is provided to the frequency analyzing unit 133.

[0019] For example, as shown in FIG. 2, the signal shaping section 132 may capture one waveform at a time by capturing the zero crossing position P3 at the rising edge of the waveform represented by the time-series signal. The position at which one waveform is captured may be the peak P1, bottom P2, or zero crossing position P4 at the falling edge of the waveform, other than the zero crossing position P3 at the rising edge. Also, one waveform may be defined and extracted by other methods.

[0020] In addition, when the time series signal contains a long-term fluctuation component in addition to the periodicity, and therefore characteristic points such as the zero crossing position P3 cannot be captured, the signal shaping unit 132 may remove the long-term fluctuation component and perform processing by calculating the difference using polynomial fitting or the difference using a moving average.

[0021] Frequency analysis unit 133 is a calculation unit that performs frequency analysis on the one waveform signal from signal shaping unit 132, and parameterizes the one waveform represented by that one waveform signal into three elements: period, power, and waveform shape. In this way, frequency analysis unit 133 acquires period parameters, power parameters, and waveform shape parameters corresponding to that one waveform. The acquired period parameters, power parameters, and waveform shape parameters are provided to parameter storage unit 134.

[0022] For example, the frequency analysis unit 133 calculates the period of one waveform or the frequency of the fundamental wave of one waveform, and performs frequency analysis processing to calculate the power parameter. The frequency analysis processing includes complex Fourier series expansion. Specifically, the frequency analysis unit 133 calculates a frequency spectrum by complex Fourier series expansion, and can normalize the frequency spectrum with the square root of the sum of squares to calculate the square root of the sum of squares, and use the resultant square root as the power parameter.

[0023] The parameter storage unit 134 stores the period parameters, power parameters, and waveform shape parameters from the frequency analysis unit 133 in the parameter storage unit 140 . The parameter storage unit 140 stores period parameters, power parameters, and waveform shape parameters.

[0024] The signal processing device 100 described above can be realized by a mobile terminal such as a smart watch or a PC (Personal Computer) corresponding to the computer 10 as shown in FIG.

[0025] The computer 10 includes a storage 11 such as a hard disk drive (HDD), a solid state drive (SSD) or a non-volatile memory, a non-volatile memory 12, a processor 13 such as a central processing unit (CPU), a sensor 14, and a connection I / F 15. The computer 10 may include a communication I / F 16 such as a network interface card (NIC) for communicating with a network.

[0026] For example, the parameter storage unit 140 can be realized by the storage 11 or the memory 12. The control unit 130 can be realized by the processor 13 loading a program stored in the storage 11 into the memory 12 and executing the program.

[0027] The detection unit 110 can be realized by the sensor 14 . The connection I / F unit 120 can be realized by a connection I / F 15 compatible with a USB (Universal Serial Bus) or the like. As described above, at least one of the sensor 14 and the connection I / F 15 may be provided. Although not shown, the computer 10 may be provided with a display interface for displaying the processing results depending on the application, so that the signal processing device 100 can be provided with a display unit (not shown) for displaying the processing results.

[0028] The program may be downloaded to the storage 11 from a recording medium (not shown) via a reader / writer (not shown) or from a network via a communication I / F (not shown), and then loaded into the memory 12 and executed by the processor 13. Alternatively, the program may be directly loaded into the memory 12 from a recording medium via a reader / writer or from a network via a communication I / F, and executed by the processor 13. In other words, the program may be provided by a computer program product such as a recording medium.

[0029] FIG. 4 is a flowchart showing the frequency analysis process in the frequency analysis unit 133. First, frequency analysis section 133 receives an input of a waveform signal from signal shaping section 132 (S10) and extracts a waveform represented by the waveform signal.

[0030] Next, the frequency analysis unit 133 calculates the waveform length from one waveform extracted in step S10 (S11). Then, the frequency analysis unit 133 acquires a period parameter from the calculated waveform length (S12). For example, the frequency analysis unit 133 may calculate a fundamental frequency by taking the reciprocal of the waveform length calculated in step S11, and use the fundamental frequency as the period parameter. Note that the waveform length itself may also be used as the period parameter.

[0031] Next, the frequency analysis unit 133 performs a frequency transformation based on a complex Fourier series expansion on the one waveform extracted in step S10 (S13). Then, the frequency analysis unit 133 calculates the real Fourier spectrum (S14). Furthermore, the frequency analysis unit 133 calculates an imaginary Fourier spectrum (S15). For example, in the complex Fourier series expansion, the frequency analysis unit 133 calculates up to n frequency components in total from the fundamental frequency to frequencies that are n (n is an integer of 2 or more) times the fundamental frequency.

[0032] Next, the frequency analysis unit 133 calculates a frequency spectrum from the square root of the sum of the squares of the frequency components of the real Fourier spectrum calculated in step S14 and the imaginary Fourier spectrum calculated in step S15 (S16). Furthermore, the frequency analysis unit 133 calculates a phase spectrum from the arctangent of each frequency component of the real Fourier spectrum calculated in step S14 and the imaginary Fourier spectrum calculated in step S15 (S17). For example, if calculations are performed up to five times the frequency, a frequency spectrum and a phase spectrum consisting of five frequency components are obtained.

[0033] Here, in step S17, frequency analysis unit 133 obtains a phase spectrum in the range of -π to π as an initial phase value in each frequency spectrum. However, while -π and π have the same meaning, they are at the farthest positions as numerical values, which may limit their use as feature quantities. Therefore, it is preferable that frequency analysis unit 133 calculates the cosine of the phase to calculate a real phase spectrum, and then calculates the sine of the phase to calculate an imaginary phase spectrum.

[0034] Next, the frequency analysis unit 133 normalizes the frequency spectrum obtained in step S16 by the square root of the sum of squares in the frequency direction (S18). As a result, the frequency analysis unit 133 calculates a normalized frequency spectrum (S19). In addition, the frequency analysis unit 133 Similar to the process in step S18 The square root of the sum of the squares is calculated (S20).

[0035] Then, the frequency analysis unit 133 acquires the normalized frequency spectrum calculated in step S19 and the phase spectrum calculated in step S17 as waveform shape parameters (S21). Note that, if the real phase spectrum and the imaginary phase spectrum are calculated in step S17, it is preferable to acquire these as waveform shape parameters.

[0036] Furthermore, the frequency analysis unit 133 acquires the square root of the sum of the squares calculated in step S20 as a power parameter (S22).

[0037] FIG. 5 is a diagram showing a schematic diagram of a waveform shape and power decomposition.

[0038] Through the above processing, the frequency analysis unit 133 realizes parameterization into the period, waveform shape, and power. As a result, the frequency analysis unit 133 can calculate the power of the waveform by taking the square root of the sum of the squares of the spectrum from the fundamental wave to the harmonic waves, thereby realizing calculation of the power in an appropriate manner that takes into account the spectrum from the fundamental wave to the harmonic waves, rather than using the amplitude value of the waveform, in other words, the peak value, as the power.Furthermore, the frequency analysis unit 133 can calculate the waveform shape separately from the strength and weakness of the entire signal.

[0039] As described above, in the first embodiment, the frequency analysis unit 133 processes the time-series signal waveform by waveform, and the parameter storage unit 113 continuously stores the processing results, so that it is possible to obtain the changes over time in parameters of the period, waveform shape, and power of the time-series signal.

[0040] Embodiment 2 FIG. 6 is a block diagram illustrating a schematic configuration of an emotion determination system 1 serving as an information processing system according to the second embodiment. The emotion determination system 1 includes a learning device 200 as an information processing device, and an inference device 300 as an information processing device. Here, the learning device 200 and the inference device 300 are capable of sending and receiving data via a network 20 such as the Internet.

[0041] The learning device 200 performs processing in a learning phase, which is a phase in which learning of a learning model is performed. Inference device 300 acquires a trained learning model from learning device 200 and performs processing in an inference phase in which emotion is detected from a pulse wave signal, which is a time-series signal, using the trained learning model. The pulse wave signal is a signal that indicates a pulse wave detected from a person.

[0042] FIG. 7 is a block diagram showing a schematic configuration of learning device 200 according to the second embodiment. The learning device 200 includes a control unit 230, a parameter storage unit 140, a communication I / F unit 250, and a learning model storage unit 260.

[0043] The parameter storage unit 140 of the learning device 200 in the second embodiment is similar to the parameter storage unit 140 of the signal processing device 100 according to the first embodiment.

[0044] The communication I / F unit 250 performs communication via the network 20 .

[0045] The control unit 230 controls the processing in the learning device 200 . The control unit 230 includes a signal acquiring unit 131 , a signal shaping unit 132 , a frequency analysis unit 133 , a parameter storage unit 134 , a data acquiring unit 235 , and a learning unit 236 .

[0046] The signal acquisition unit 131, the signal shaping unit 132, the frequency analysis unit 133 and the parameter storage unit 134 of the control unit 230 in embodiment 2 are similar to the signal acquisition unit 131, the signal shaping unit 132, the frequency analysis unit 133 and the parameter storage unit 134 of the control unit 130 in embodiment 1. However, the signal acquiring unit 131 acquires a time-series signal from the data acquiring unit 235, as described later.

[0047] Data acquisition unit 235 acquires teacher data including a pulse wave signal as a time-series signal and a label indicating the emotion of the subject from whom the pulse wave signal was acquired. Here, the label indicates the emotion of the subject at each predetermined period in the pulse wave signal. Note that here, the subject from whom the pulse wave signal is acquired is a person, and this person is also referred to as a learning subject.

[0048] Data acquisition unit 235 may acquire teacher data from another device (not shown), for example, via communication I / F unit 250, or may acquire the teacher data by reading the teacher data from a storage unit (not shown). The pulse wave signal included in the acquired teacher data is provided to signal acquisition unit 131, and the label is provided to learning unit 236.

[0049] Learning unit 236 learns a learning model for inferring, from a pulse wave signal, the emotion of the subject from whom the pulse wave signal was acquired, by using the period parameters, power parameters, and waveform shape parameters for each predetermined period stored in parameter storage unit 140, and the labels corresponding to the periods. Here, the power parameters, period parameters, and waveform shape parameters used for learning are also referred to as learning power parameters, learning period parameters, and learning waveform shape parameters, respectively.

[0050] For example, the learning unit 236 uses at least the power parameters as examples and learns the person's emotions when a waveform corresponding to the power parameters is acquired as a correct answer, thereby generating a learning model for inferring the person's emotions. Note that a waveform identified from the time-series signal used for learning is also referred to as a learning waveform.

[0051] As the machine learning method used by the learning unit 236, a neural network or a support vector machine can be selected as a good example, but other machine learning methods may be used. As the parameters used for learning, at least one of the period parameter, the power parameter, and the waveform shape parameter may be used. Here, it is assumed that at least the power parameter is used.

[0052] The learning model storage unit 260 stores the learning model learned by the learning unit 236. The communication I / F unit 250 transmits the learning model stored in the learning model memory unit 260 to the inference device 300.

[0053] The above-described learning device 200 can also be realized by a computer 10 as shown in FIG. For example, the parameter storage unit 140 and the learning model storage unit 260 can be realized by the storage 11 or the memory 12. The control unit 230 can be realized by the processor 13 loading a program stored in the storage 11 into the memory 12 and executing the program. The communication I / F unit 250 can be realized by the communication I / F 16 .

[0054] FIG. 8 is a block diagram showing a schematic configuration of an inference device 300 according to the second embodiment. The inference device 300 includes a detection unit 110, a connection I / F unit 120, a control unit 330, a parameter storage unit 140, a communication I / F unit 350, and a learning model storage unit 360.

[0055] The detection unit 110, the connection I / F unit 120 and the parameter storage unit 140 of the inference device 300 in the second embodiment are similar to the detection unit 110, the connection I / F unit 120 and the parameter storage unit 140 of the signal processing device 100 in the first embodiment.

[0056] The communication I / F unit 350 performs communication via the network 20 . For example, the communication I / F unit 350 receives a learning model from the learning device 200 via the network 20. The learning model storage unit 360 stores the learning model received by the communication I / F unit 350.

[0057] The control unit 330 controls the processing in the inference device 300 . The control unit 330 includes a signal acquisition unit 131 , a signal shaping unit 132 , a frequency analysis unit 133 , a parameter storage unit 134 , and an inference unit 337 . The signal acquisition unit 131, the signal shaping unit 132, the frequency analysis unit 133 and the parameter storage unit 134 of the control unit 330 in embodiment 2 are similar to the signal acquisition unit 131, the signal shaping unit 132, the frequency analysis unit 133 and the parameter storage unit 134 of the control unit 130 in embodiment 1.

[0058] The inference unit 337 inputs the period parameters, power parameters, and waveform shape parameters for each predetermined period stored in the parameter storage unit 140 into the learning model stored in the learning model storage unit 360, and infers the emotion corresponding to the parameters. The parameters used by the inference unit 337 may be the parameters used by the learning device 200 for learning.

[0059] The above-described inference apparatus 300 can also be realized by a computer 10 as shown in FIG. For example, the parameter storage unit 140 and the learning model storage unit 360 can be realized by the storage 11 or the memory 12. The control unit 330 can be realized by the processor 13 loading a program stored in the storage 11 into the memory 12 and executing the program. The communication I / F unit 350 can be realized by the communication I / F 16 .

[0060] For example, the emotion determination system 1 according to the second embodiment may be applied when classifying emotions into pleasure and discomfort, or into awakening and non-awakening, using a data set of pulse wave signals labeled with emotional states such as pleasure and discomfort, or awakening and non-awakening.

[0061] It is known that emotions, which are internal states of a person, are expressed in the pulse waveform (waveform shape) that represents changes within a single waveform, and it is also possible to use only the parameters of the pulse waveform (waveform shape) for multiple pulses and perform learning using a CNN (Convolutional Neural Network).By doing this, it is possible to separate the components related to the strength of the entire signal by separating the power from the pulse waveform, and to properly learn the pulse waveform components consisting of the biomechanics of blood vessels in the time-frequency domain, enabling proper classification.

[0062] Embodiment 3 In the second embodiment described above, classification is performed by machine learning based on parameters, but waveform prediction may also be performed by performing regression. As shown in FIG. 6, a signal prediction system 2 as an information processing system according to the third embodiment includes a learning device 400 as an information processing device, and an inference device 500 as an information processing device. Here, learning device 400 and inference device 500 are capable of sending and receiving data via network 20 such as the Internet.

[0063] The learning device 400 performs processing in a learning phase, which is a phase in which learning of a learning model is performed. The inference device 500 acquires a trained learning model from the learning device 400 and performs processing in the inference phase, which is a phase in which future time-series signals are predicted from time-series signals using the trained learning model.

[0064] FIG. 9 is a block diagram showing a schematic configuration of a learning device 400 according to the third embodiment. The learning device 400 includes a detection unit 110, a connection I / F unit 120, a control unit 430, a parameter storage unit 440, a communication I / F unit 250, and a learning model storage unit 260.

[0065] The detection unit 110 and the connection I / F unit 120 of the learning device 400 in the fourth embodiment are similar to the detection unit 110 and the connection I / F unit 120 of the signal processing device 100 in the first embodiment. Furthermore, the communication I / F unit 250 and the learning model storage unit 260 of the learning device 400 in the fourth embodiment are similar to the communication I / F unit 250 and the learning model storage unit 260 of the learning device 200 in the second embodiment.

[0066] The control unit 430 controls the processing in the learning device 400 . The control unit 430 includes a signal acquisition unit 131 , a signal shaping unit 132 , a frequency analysis unit 133 , a parameter storage unit 434 , and a learning unit 436 .

[0067] The signal acquiring section 131, the signal shaping section 132 and the frequency analyzing section 133 of the control section 430 in the fourth embodiment are similar to the signal acquiring section 131, the signal shaping section 132 and the frequency analyzing section 133 of the control section 130 in the first embodiment. However, the signal acquiring unit 131 also provides the acquired time-series signal to the parameter storage unit 434 .

[0068] The parameter storage unit 434 stores the period parameters, power parameters, and waveform shape parameters from the frequency analysis unit 133 and the time-series signal in the parameter storage unit 440 .

[0069] The parameter storage unit 440 stores period parameters, power parameters, waveform shape parameters, and time-series signals.

[0070] The learning unit 436 uses the period parameters, power parameters, and waveform shape parameters for each predetermined period stored in the parameter storage unit 140 to learn a learning model for predicting a time-series signal in the future.

[0071] For example, the learning unit 436 uses at least the power parameters corresponding to the time series signal at a first time as an example, and learns a time series signal at a second time that is later than the first time as a correct answer, thereby generating a learning model for inferring a future time series signal. Note that the power parameters, period parameters, and waveform shape parameters used for learning are also referred to as learning power parameters, learning period parameters, and learning waveform shape parameters, respectively.

[0072] As the machine learning method used by the learning unit 436, a neural network is a good example, but other machine learning methods may be used. At least one of the period parameter, power parameter, and waveform shape parameter may be used as the parameter used for learning. Here, it is assumed that at least the power parameter is used.

[0073] The learning model learned in the above manner is stored in the learning model storage unit 260 and transmitted to the inference device 500 via the communication I / F unit 250.

[0074] The above-described learning device 400 can also be realized by a computer 10 as shown in FIG. For example, the parameter storage unit 440 and the learning model storage unit 260 can be realized by the storage 11 or the memory 12. The control unit 430 can be realized by the processor 13 loading a program stored in the storage 11 into the memory 12 and executing the program. The communication I / F unit 250 can be realized by the communication I / F 16 .

[0075] As shown in FIG. 8, inference device 500 in embodiment 3 includes a detection unit 110, a connection I / F unit 120, a control unit 530, a parameter memory unit 140, a communication I / F unit 350, and a learning model memory unit 360.

[0076] The detection unit 110, the connection I / F unit 120 and the parameter storage unit 140 of the inference device 500 in the third embodiment are similar to the detection unit 110, the connection I / F unit 120 and the parameter storage unit 140 of the signal processing device 100 in the first embodiment. The communication I / F unit 350 and the learning model storage unit 360 of the inference device 500 in the third embodiment are similar to the communication I / F unit 350 and the learning model storage unit 360 of the inference device 300 in the second embodiment.

[0077] The control unit 530 controls the processing in the inference device 500 . The control unit 530 includes a signal acquisition unit 131 , a signal shaping unit 132 , a frequency analysis unit 133 , a parameter storage unit 134 , and an inference unit 537 . The signal acquisition unit 131, the signal shaping unit 132, the frequency analysis unit 133 and the parameter storage unit 134 of the control unit 530 in embodiment 3 are similar to the signal acquisition unit 131, the signal shaping unit 132, the frequency analysis unit 133 and the parameter storage unit 134 of the control unit 130 in embodiment 1.

[0078] The inference unit 537 inputs the period parameters, power parameters, and waveform shape parameters for each predetermined period stored in the parameter memory unit 140 into the learning model stored in the learning model memory unit 360, and infers a future time series signal from the parameters. The parameters used by the inference unit 537 may be the parameters used by the learning device 400 for learning.

[0079] The above-described inference apparatus 500 can also be implemented by a computer 10 as shown in FIG. For example, the parameter storage unit 140 and the learning model storage unit 360 can be realized by the storage 11 or the memory 12. The control unit 530 can be realized by the processor 13 loading a program stored in the storage 11 into the memory 12 and executing the program. The communication I / F unit 350 can be realized by the communication I / F 16 .

[0080] For example, the signal processing described in the third embodiment may be applied when predicting future changes in daily average temperatures in Tokyo from data for about 10 years that stores changes in daily average temperatures in Tokyo. In this case, the waveform shape represents the manner in which the seasons change in Tokyo, the power represents the degree of change in the strength of temperature differences throughout the year, and the period represents the period of the seasons. Here, since the period does not change over a one-year period, period parameters are not necessary for learning. Therefore, by using the two parameters of the waveform shape and power to perform learning using LSTM (Long Short Term Memory), it is possible to perform inference more appropriately than with conventional parameterization and learning methods. In this case, for example, a learning model may be trained to predict the time series signal for the next year from parameters for 10 years.

[0081] Embodiment 4 FIG. 10 is a block diagram showing a schematic configuration of a signal reconstructing device 600 which is an information processing device according to the fourth embodiment. The signal reconstruction device 600 includes a detection unit 110 , a connection I / F unit 120 , a control unit 630 , and a parameter storage unit 140 . Although both the detection unit 110 and the connection I / F unit 120 are provided in FIG. 10, the signal reconstruction device 600 only needs to include at least one of them.

[0082] The detection unit 110, the connection I / F unit 120 and the parameter storage unit 140 of the signal reconstruction device 600 of embodiment 4 are similar to the detection unit 110, the connection I / F unit 120 and the parameter storage unit 140 of the signal processing device 100 of embodiment 1.

[0083] The control unit 630 controls the processing in the signal reconstruction device 600 . The control unit 630 includes a signal acquisition unit 131 , a signal shaping unit 132 , a frequency analysis unit 133 , a parameter storage unit 134 , a signal filter unit 638 , and an inverse conversion unit 639 .

[0084] The signal acquisition unit 131, the signal shaping unit 132, the frequency analysis unit 133 and the parameter storage unit 134 of the control unit 630 in embodiment 6 are similar to the signal acquisition unit 131, the signal shaping unit 132, the frequency analysis unit 133 and the parameter storage unit 134 of the control unit 130 in embodiment 1.

[0085] The signal filter unit 638 performs filtering on at least one of the power parameter, the period parameter, and the waveform shape parameter. For example, the signal filter unit 638 executes a process of deleting unnecessary components from among the period parameters, power parameters, and waveform shape parameters stored in the parameter storage unit 140, or replacing them with constants.

[0086] Specifically, when eliminating periodic fluctuations in a time series signal, the signal filter unit 638 replaces the period parameter with a constant. Moreover, when eliminating the change in strength over time of the time series signal, the signal filter unit 638 replaces the power parameter with a constant. Furthermore, when eliminating harmonic components of the time series signal and expressing it as a simple sine wave, the signal filter unit 638 replaces the harmonic components of the waveform shape parameters with zero. The parameters filtered in this manner are provided to an inverse transform unit 639 .

[0087] The inverse conversion unit 639 converts the power parameters, period parameters, and waveform shape parameters after the filtering process by the signal filter unit 638 into time-series signals. For example, the inverse transform unit 639 uses the parameters filtered by the signal filter unit 638 to perform inverse complex Fourier series expansion for each waveform, thereby calculating a reconstructed signal.

[0088] For example, in a pulse wave signal obtained as a time series signal by a pulse wave Doppler sensor, the distance from the signal provider affects the amplitude of the pulse wave signal, and the information on the strength of each wave of the waveform may be regarded as noise. In this case, a reconstructed signal from which the strength components of the signal over time have been removed can be output by using the pulse period and pulse waveform parameters to replace the pulse wave power with a constant that does not change for each waveform and then performing an inverse transform.

[0089] The above-described signal reconstruction device 600 can also be realized by a computer 10 as shown in FIG. For example, the control unit 630 can be realized by the processor 13 loading a program stored in the storage 11 into the memory 12 and executing the program. [Explanation of symbols]

[0090] 1 emotion determination system, 2 signal prediction system, 100 signal processing device, 200,400 learning device, 300,500 inference device, 600 signal reconstruction device, 110 detection unit, 120 connection I / F unit, 130,230,330,430,530,630 control unit, 131 signal acquisition unit, 132 signal shaping unit, 133 frequency analysis unit, 134,434 parameter storage unit, 235 data acquisition unit, 236,436 learning unit, 337,537 inference unit, 638 signal filter unit, 639 inverse transformation unit, 140,440 parameter memory unit, 250,350 communication I / F unit, 260,360 learning model memory unit.

Claims

1. a signal acquisition unit that acquires a time-series signal indicating a plurality of waveforms; an identifying unit that identifies one waveform from the time-series signal; a calculation unit that calculates a power parameter that is a parameter indicating the power of the one waveform, a period parameter that is a parameter indicating the period of the one waveform, and a waveform shape parameter that is a parameter indicating the waveform shape obtained by separating the power component from the one waveform. An information processing device characterized by:

2. The calculation unit calculates the period of the one waveform or the fundamental frequency of the one waveform and performs frequency analysis processing to calculate the power parameter, the period parameter, and the waveform shape parameter.

2. The information processing device according to claim 1,

3. The frequency analysis process includes a complex Fourier series expansion.

3. The information processing device according to claim 2, wherein:

4. the calculation unit calculates a frequency spectrum by the complex Fourier series expansion, normalizes the frequency spectrum by the square root of the sum of squares, and sets the square root of the sum of squares calculated as the power parameter, the fundamental wave frequency as the period parameter, and sets the spectrum normalized by the square root of the sum of squares as the waveform shape parameter.

4. The information processing device according to claim 3,

5. the time-series signal is a signal indicating a pulse wave detected from a person; The apparatus further includes a learning unit that generates a learning model for inferring the human emotion by learning the human emotion at the time when the one waveform is acquired as a correct answer using the power parameters as an example. The information processing device according to claim 1 ,

6. the time-series signal is a signal indicating a pulse wave detected from a person; The device further includes an inference unit that uses a learning power parameter, which is a parameter indicating the power of a learning waveform that is a waveform identified from a signal indicating a pulse wave detected from a person to be trained, as an example, and inputs the power parameter into a learning model that is generated by learning the emotion of the person to be trained when the learning waveform is acquired as a correct answer. The information processing device according to claim 1 ,

7. The apparatus further includes a learning unit that generates a learning model for inferring future time series signals by using the power parameters corresponding to the time series signals at a first time as an example and learning a time series signal at a second time that is later than the first time as a correct answer. The information processing device according to claim 1 ,

8. The apparatus further includes an inference unit that uses a learning power parameter indicating the power of one waveform identified from a time-series signal indicating a plurality of waveforms at a first time as an example, and inputs the power parameter into a learning model that is generated by learning a time-series signal indicating a waveform at a second time that is later than the first time as a correct answer, thereby inferring a time-series signal in the future. The information processing device according to claim 1 ,

9. a signal filter unit that performs filtering on at least one of the power parameter, the period parameter, and the waveform shape parameter; an inverse conversion unit that converts the power parameters, period parameters, and waveform shape parameters after the filtering process into time-series signals. The information processing device according to claim 1 ,

10. Computer, a signal acquisition unit that acquires a time-series signal showing a plurality of waveforms; an identifying unit that identifies one waveform from the time-series signal; and a calculation unit that calculates a power parameter that is a parameter indicating the power of the one waveform, a period parameter that is a parameter indicating the period of the one waveform, and a waveform shape parameter that is a parameter indicating the waveform shape obtained by separating the power component in the one waveform. A program characterized by.

11. Acquire a time series signal showing multiple waveforms, Identifying a waveform from the time series signal; Calculating a power parameter which is a parameter indicating the power of the one waveform, a period parameter which is a parameter indicating the period of the one waveform, and a waveform shape parameter which is a parameter indicating the waveform shape obtained by separating the power component in the one waveform. An information processing method comprising:

Images