Information processing device and information processing method

Abstract

An information processing device according to the present disclosure comprises: a feature amount extraction unit that extracts, from a biological signal of a user, a physiological reaction which contributes to the emotion of the user and which serves as a feature amount; a correction gain calculation unit that calculates, on the basis of the extracted feature amount, a correction gain for correcting an individual difference in the physiological reaction accompanying an emotional change of the user; and a normalization unit that normalizes the extracted feature amount on the basis of a prescribed normalization coefficient and the calculated correction gain.

Description

Information processing device and information processing method

[0001] The present disclosure relates to an information processing device and an information processing method that perform emotion estimation.

[0002] Changes in a person's emotions are expressed on the body surface as physiological responses such as brain waves, heart rate, and sweating. To estimate a person's emotions, these physiological responses are read as biosignals by a sensor device, feature quantities such as physiological indices that contribute to the emotional response are extracted by signal processing, and the user's emotions are estimated from the feature quantities using a model formula obtained by machine learning (see Patent Documents 1 and 2).

[0003] JP 2020-130438 A JP 2023-89729 A

[0004] Even when performing concentrated work or experiencing the same level of stress, there are individual differences in the physiological responses, which leads to individual differences in the output of the emotion estimation model. When estimating emotions, it is desirable to adjust the physiological response range to suit each individual.

[0005] Therefore, it is desirable to provide an information processing device and an information processing method that are capable of highly accurate emotion estimation that is less affected by individual differences in physiological responses.

[0006] An information processing device according to one embodiment of the present disclosure includes a feature extraction unit that extracts physiological responses that contribute to a user's emotions as features from the user's biosignals, a correction gain calculation unit that calculates a correction gain from the extracted feature to correct individual differences in physiological responses associated with changes in the user's emotions, and a normalization unit that normalizes the extracted feature based on a predetermined normalization coefficient and the calculated correction gain.

[0007] An information processing method according to one embodiment of the present disclosure includes extracting, from a user's biosignal, a physiological response that contributes to the user's emotion as a feature; calculating, from the extracted feature, a correction gain for correcting individual differences in the physiological response associated with changes in the user's emotion; and normalizing the extracted feature based on a predetermined normalization coefficient and the calculated correction gain.

[0008] In an information processing device or information processing method according to an embodiment of the present disclosure, physiological responses that contribute to a user's emotions are extracted as features, a correction gain is calculated from the extracted features to correct individual differences in physiological responses associated with changes in the user's emotions, and the extracted features are normalized based on a predetermined normalization coefficient and the calculated correction gain.

[0009] FIG. 1 is a block diagram schematically illustrating an example configuration of an information processing device according to an embodiment of the present disclosure. FIG. 2 is a flowchart illustrating a specific example of a behavioral state determination process performed by a behavioral state determination unit in the information processing device according to the embodiment. FIG. 3 is a flowchart illustrating an example of a correction gain calculation process performed by an emotional response range correction gain calculation unit in the information processing device according to the embodiment. FIG. 4 is an explanatory diagram illustrating an example of a correction gain calculation process performed by the emotional response range correction gain calculation unit when the amount of data of accumulated measured biological signals is less than a certain level. FIG. 5 is an explanatory diagram illustrating an example of a correction gain calculation process performed by the emotional response range correction gain calculation unit when the amount of data of accumulated measured biological signals is equal to or greater than a certain level. FIG. 6 is a block diagram schematically illustrating an example configuration of an information processing device according to a modified example. FIG. 7 is an explanatory diagram illustrating a modified example of a correction gain calculation process performed by the emotional response range correction gain calculation unit when the amount of data of accumulated measured biological signals is equal to or greater than a certain level.

[0010] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. The description will be made in the following order: 0. Comparative Example 1. One Embodiment 1.1 Configuration and Operation 1.2 Specific Example of Correction Gain Calculation 1.3 Modification 1.4 Effects 2. Other Embodiments

[0011] <0. Comparative Example> (Overview and Issues of an Information Processing Device According to a Comparative Example) Changes in a person's emotions are expressed on the body surface as physiological responses such as brain waves, heart rate, and sweating. To estimate a person's emotions, these physiological responses are read as biosignals using a sensor device, and feature quantities such as physiological indicators that contribute to the emotional response are extracted through signal processing. The user's emotions are then estimated from the feature quantities using a model formula obtained through machine learning. Types of emotions are classified along two axes: pleasant / unpleasant and arousal, as described in the following academic literature 1, for example. [Academic literature 1] Posner, Jonathan, James A. Russell, and Bradley S. Peterson. "The circumplex model of affect: An integrative approach to affective neuroscience, cognitive development, and psychopathology." Development and psychopathology 17.3 (2005): 715-734.

[0012] When constructing a physiologically valid emotion estimation model in a laboratory environment and applying it to real-world problems, individual differences in baseline states of human emotional physiological responses become an issue. Furthermore, human physiological responses involve sleep-wake rhythms. Sleep-wake rhythms are regulated to approximately daily rhythms by the body clock, similar to the autonomic nervous system (e.g., body temperature), endocrine hormone system, and immune / metabolic system. Rhythms with a period of approximately one day like this are called circadian rhythms.

[0013] The following academic paper discusses the importance of measuring resting baseline conditions in psychological experiments: [Academic Paper 2] Laborde, Sylvain, et al., "Heart rate variability and cardiac vagal tone in psychophysiological research—recommendations for experiment planning, data analysis, and data reporting." Frontiers in psychology 8 (2017): 213.

[0014] The following academic literature 3 is related to the hypothesis of the cardiac vagal nerve's physiological response (vagal tank theory). The vagal tank theory classifies emotional changes resulting from the physiological response of the cardiac vagus nerve into instantaneous changes (phasic) and gradual changes (tonic). A person's resting state serves as the baseline, and the subsequent state is determined by external stimuli (events). The vagal tank theory uses a tank metaphor to represent the balance between stress and relaxation, depending on whether the external stimulus is a stressful or relaxing event. [Academic literature 3] Laborde, Sylvain, et al., "Vagal tank theory: the three Rs of cardiac vagal control functioning—resting, reactivity, and recovery." Frontiers in neuroscience 12 (2018): 458.

[0015] The following academic paper, Reference 4, discusses the neural circuitry related to the relationship between human wakefulness and circadian rhythms. [Academic Reference 4] Aston-Jones, Gary, et al. "A neural circuit for circadian regulation of arousal." Nature neuroscience 4.7 (2001): 732-738.

[0016] When estimating emotions in real environments, particularly arousal levels such as stress state estimation and concentration / relaxation estimation, understanding the physiological and neurological characteristics shown in the above academic literature can address individual differences in human physiological responses and improve the accuracy of emotion estimation. A particular challenge in practical applications is the issue of individual differences in physiological responses even when performing the same level of concentration work or stress load. These individual differences in physiological responses can also result in individual differences in the output of the emotion estimation model, making it necessary to adjust the range of variation to suit each individual. To achieve high emotion estimation accuracy depending on the application, the system must be able to control the individual differences in physiological responses shown in the above literature.

[0017] Patent Document 1 (JP 2020-130438 A) proposes a technology for generating a circadian rhythm based on a user's biological information and behavioral state, and predicting and estimating biological signals. The technology described in Patent Document 1 requires generating a circadian rhythm once, assuming continuous measurement, which can limit the user experience when using a sensor device that does not assume continuous measurement, such as earphones or headphones.

[0018] Patent Document 2 (JP 2023-89729 A) proposes a technology for generating a biological signal time series and a motion signal time series, and correcting the biological signal time series using the motion signal time series to a corrected biological signal time series in which the influence of the user's motion is reduced. The technology described in Patent Document 2 has difficulty correcting individual variations in physiological responses according to circadian rhythms.

[0019] The technologies described in Patent Documents 1 and 2 are based on the premise of continuous long-term measurement data using sensor devices such as wristbands or chest-mounted sensors. Therefore, it is difficult to directly apply them to short-term measurements using sensor devices that are frequently removed, such as earphones. If it were possible to correct individual differences in physiological responses in response to emotional changes without requiring users to wear sensor devices for long periods of time, it would be possible to improve the accuracy of emotion estimation in a variety of everyday situations, and the expansion of applications using emotion estimation is expected.

[0020] 1. One Embodiment> [1.1 Configuration and Operation] FIG. 1 schematically illustrates an example configuration of an information processing device according to one embodiment of the present disclosure.

[0021] The information processing device according to one embodiment includes an emotion estimation unit 10 .

[0022] The emotion estimation unit 10 includes a feature extraction unit 11, a behavioral state determination unit 12, a signal quality determination unit 13, an emotional response range correction gain calculation unit 14, a feature normalization unit 15, and an emotional state determination unit 16. The emotion estimation unit 10 further includes a normalization coefficient storage unit 21, a user attribute storage unit 22, and a model parameter storage unit 23.

[0023] An information processing device according to an embodiment may be configured as a computer including, for example, one or more central processing units (CPUs), one or more read-only memories (ROMs), and one or more random access memories (RAMs). In this case, the processing of each block in the information processing device according to an embodiment may be realized by one or more CPUs executing processing based on a program stored in one or more ROMs or RAMs. Furthermore, the processing of each block in the information processing device according to an embodiment may be realized by one or more CPUs executing processing based on a program supplied from an external source via, for example, a wired or wireless network.

[0024] The emotion estimation unit 10 is an estimation unit that estimates the emotional state of a user whose emotional state is to be estimated based on a measurement signal from the sensor device 30.

[0025] The sensor device 30 includes a vital sensor (biometric sensor) and a physical sensor such as a motion sensor. The sensor device 30 may be one or more devices. The biometric sensor and the physical sensor may be mounted on a single device or on separate devices. The sensor device 30 may be a mobile device such as a smartphone equipped with a biometric sensor or a physical sensor, or various wearable devices. The physical sensor may include a global positioning system (GPS), an inertial measurement unit (IMU), a microphone, etc. The biometric sensor detects biosignals for estimating the user's emotional state, such as sweating, pulse wave (PPG), electromyography (EMG), blood pressure, blood flow, or body temperature. The emotion estimation unit 10 estimates the user's emotional state based on the biometric signal detected by the biometric sensor. This emotional state can be used to determine the user's concentration state, wakefulness state, etc.

[0026] The sensor device 30 outputs sensor information from a biosensor and sensor information from a physical sensor. The sensor device 30 outputs the sensor information from the biosensor as vital information (biological information) to the feature extractor 11 and the signal quality determiner 13. The sensor device 30 also outputs the sensor information from the physical sensor as physical information to the behavioral state determiner 12. If the sensor device 30 includes a smartphone, the sensor device 30 may also output information such as the smartphone's application usage history as physical information to the behavioral state determiner 12.

[0027] The feature extraction unit 11 is an extraction unit that extracts a physiological response that contributes to an emotion from a biosignal as a feature x. The feature extraction unit 11 extracts a physiological response that contributes to an emotion of a user from a biosignal obtained from a biosensor of the sensor device 30 as a feature x. The feature extraction unit 11 extracts a physiological response that contributes to an emotion of a user from a biosignal obtained from a biosensor of the sensor device 30 as a feature x. The feature extraction unit 11 extracts a feature x=(x １ , x ２ , ..., x Ｎ The feature extraction unit 11 outputs the extracted feature x to the feature normalization unit 15 and the emotional response range correction gain calculation unit 14.

[0028] The feature extraction unit 11 observes biological signals from vital sensors (biosensors) such as electroencephalogram (EEG), psychoacoustic dehydration (EDA), pulse wave (PPG), or blood flow (LDF) as time-series data, and extracts physiological indices that contribute to changes in emotions as feature values ​​x. Typically, feature values ​​x are extracted using an analysis window (sliding window) of about 5 minutes.

[0029] The behavioral state determination unit 12 determines the behavioral state of the user based on physical information. The behavioral state determination unit 12 determines the behavior of a user holding a sensor device 30, for example, a mobile device, based on sensor information (physical information) from a physical sensor mounted on the device. The behavioral state determination unit 12 determines the behavioral state (context) by, for example, pattern recognition of waveform data from a motion sensor as the sensor device 30 to determine the type of movement or the type of vehicle, but the determination method is not limited thereto. The behavioral state determination unit 12 may determine various modes of transportation, such as bicycle, train, car, or elevator, in addition to movements such as walking and running, without using positioning information. The behavioral state determination unit 12 may also output information such as smartphone application usage as context information. The behavioral state determination unit 12 may determine whether the user is in an active state or an inactive state as the behavioral state.

[0030] FIG. 2 is a flowchart showing a specific example of a behavioral state determination process performed by the behavioral state determination unit 12 of the information processing device according to an embodiment.

[0031] First, the behavioral state determination unit 12 acquires sensor data from the sensor device 30 (step S301). Next, the behavioral state determination unit 12 extracts a feature value x from the sensor data from the sensor device 30 (step S302). Next, the behavioral state determination unit 12 performs pattern recognition using machine learning (step S303). Next, the behavioral state determination unit 12 determines movement and vehicle type as behavior recognition results (step S304).

[0032] The signal quality determination unit 13 determines the signal quality of the measured biosignal. The signal quality determination unit 13 analyzes the waveform of the biosignal measured by the biosensor of the sensor device 30, and detects and distinguishes the type of artifact (such as noise other than the target signal) for each waveform of the biosignal using a machine learning model. The signal quality determination unit 13 can also detect noise waveforms from which artifact removal is difficult. The signal quality determination unit 13 determines the signal quality based on the determination result and calculates a signal quality score.

[0033] The emotional response range correction gain calculation unit 14 is a correction gain calculation unit that calculates a correction gain Ga for correcting individual differences in physiological responses accompanying emotional changes of the user from the feature amount x extracted by the feature amount extraction unit 11. The emotional response range correction gain calculation unit 14 switches the calculation method of the correction gain Ga depending on the amount of data of the accumulated measured biological signals of the user.

[0034] The emotional response range correction gain calculation unit 14 has a data accumulation unit 14A. The data accumulation unit 14A accumulates data on the measured biological signals and data on the value of the correction gain Ga. In an information processing device according to one embodiment, the method for calculating the correction gain Ga is switched depending on the amount of data on the measured biological signals of the user accumulated by the data accumulation unit 14A, and correction of the correction gain Ga is performed in stages. Note that the data accumulation unit 14A may not be provided inside the emotional response range correction gain calculation unit 14, but may be provided in a location separate from the emotional response range correction gain calculation unit 14.

[0035] (Step 1) When the amount of data of the user's actual measured biological signals accumulated by the data accumulation unit 14A is less than a certain level: the emotional response range correction gain calculation unit 14 calculates the fluctuation range of the feature amount x corresponding to the fluctuation range of the user's physiological response based on the user's user attributes, and calculates the correction gain Ga based on the calculated fluctuation range of the feature amount x. Information on the user attributes is stored in advance in the user attribute storage unit 22. (Step 2) When the amount of data of the user's actual measured biological signals accumulated by the data accumulation unit 14A is equal to or greater than a certain level: the emotional response range correction gain calculation unit 14 utilizes the characteristics of the circadian rhythm to estimate the fluctuation range of the feature amount x corresponding to emotional changes from short-term actual measurement data. The circadian rhythm is a rhythm of alertness according to the time of day, and has characteristics such as low alertness when waking up and high alertness after noon. Specifically, the emotional response range correction gain calculation unit 14 calculates a fluctuation range of the feature amount x corresponding to the fluctuation range of the physiological response based on data on the feature amount x accumulated over multiple days during at least the time period in which the user's emotion is estimated, and calculates the correction gain Ga based on the calculated fluctuation range of the feature amount x. (Stage 3) In (Stage 2), the emotional response range correction gain calculation unit 14 updates the value of the correction gain Ga for each day depending on the degree of accumulation of data on the value of the correction gain Ga for each day, and stabilizes the value of the correction gain Ga.

[0036] In the above (Step 2), if there are multiple time periods during which the amount of accumulated data on the user's actual measured biological signals is equal to or greater than a certain level, the emotional response range correction gain calculation unit 14 may calculate a representative value of the correction gain Ga for each day based on the data on the feature amount x for each time period. Also, the emotional response range correction gain calculation unit 14 may calculate a representative value of the correction gain Ga for each day based on a weighted average of the fluctuation range of the feature amount x for each time period based on the reliability of the fluctuation range of the feature amount x for each time period.

[0037] The normalization coefficient storage unit 21 stores a predetermined normalization coefficient (σ ｍｏｄｅｌ The feature normalization unit 15 stores a predetermined normalization coefficient (σ ｍｏｄｅｌ) and the correction gain Ga calculated by the emotional response range correction gain calculation unit 14. The normalization method used by the feature extraction unit 11 includes MIN-MAX normalization (normalization to 0 to 1 based on MAX-MIN of the distribution of the feature x) and Z standardization (standardization based on the mean and variance of the distribution of the feature x).

[0038] The feature normalization unit 15 normalizes the feature x based on the correction gain Ga so that the scale of the variation range of the feature x during model training of the emotion estimation model and the variation range of the feature x during emotion estimation in the real environment are aligned. This allows the emotional state determination unit 16 to determine the user's emotional state in which individual differences in physiological responses accompanying emotional changes have been corrected.

[0039] During model learning, the feature normalization unit 15 normalizes, for example, the fluctuation range σ of the feature x of the data during model learning. ｍｏｄｅｌ The feature quantity x is scaled (normalized) based on the normalization coefficient σ. The normalized feature quantity Z is calculated, for example, by the following formula (A): ｍｏｄｅｌ indicates the fluctuation range of the feature value x during model learning, and the x bar indicates the reference value of the feature value x. The reference value of the feature value x is the minimum value x of the feature value x. ｍｉｎ , or in a modified example described later, the baseline feature x ｂａｓｅｌｉｎｅ Shows.

[0040]

[0041] When estimating emotions in a real environment, the feature normalization unit 15 normalizes the fluctuation range σ of the feature x associated with emotional changes in the dataset during model learning, as shown in the following formula (B). ｍｏｄｅｌ The correction gain Ga is multiplied by the feature value x during model learning, and the fluctuation range σ ｍｏｄｅｌ The feature x is scaled (normalized) so that it is consistent with the fluctuation range σ (denoted as σ(^)) of the feature x during emotion estimation in the real environment, where σ(^) represents the fluctuation range of the feature x in the data during inference in the real environment, x bar represents the reference value of the feature x, and Ga represents the correction gain.

[0042]

[0043] The model parameter storage unit 23 stores model parameters of a machine learning model used for emotion estimation. The emotional state determination unit 16 determines the emotional state based on the normalized feature amount x corrected according to individual differences by the feature amount normalization unit 15. This enables highly accurate emotion estimation that is less affected by individual differences in physiological responses. The emotional state determination unit 16 determines the emotional state based on the machine learning model.

[0044] The emotional state determination unit 16 analyzes the pattern of the time-series data of the normalized feature quantity x calculated by the feature quantity normalization unit 15, and determines the emotional state of the user. ｌ = (x １ , x ２ , ..., xl) is used as input, and a sequence of emotional states is calculated as Y based on a pre-constructed machine learning model. ｌ = (y １ , y ２ , ..., yl). The emotional state determination unit 16 estimates, for example, arousal level as the emotional state. The emotional state determination unit 16 outputs, as the arousal level estimation result, a value (for example, 0.0 to 1.0) that increases as the arousal level increases.

[0045] [1.2 Specific Example of Correction Gain Calculation] FIG. 3 is a flowchart showing an example of a correction gain calculation process performed by the emotional response range correction gain calculation unit 14 in the information processing device according to one embodiment.

[0046] First, the emotional response range correction gain calculation unit 14 determines whether the amount of data of the measured biological signals stored in the data storage unit 14A is equal to or greater than a certain level (step S400). If the amount of data of the measured biological signals stored is less than the certain level (step S400; N), the emotional response range correction gain calculation unit 14 calculates a correction gain Ga based on the user attributes previously stored in the user attribute storage unit 22 (step S401).

[0047] FIG. 4 shows an example of the correction gain calculation process by the emotional response range correction gain calculation unit 14 when the amount of accumulated data of the measured biological signals does not reach a certain level.

[0048] When the amount of data of the measured biological signals stored in the data storage unit 14A is less than a certain level, the emotional response range correction gain calculation unit 14 calculates the fluctuation range σ(^) of the user's feature quantity corresponding to the fluctuation range of the user's physiological response based on the user attributes and using statistics of the maximum and minimum values ​​of the feature quantity x stored in advance. Specifically, the fluctuation range σ(^) is calculated, for example, by the following formula (1).

[0049]

[0050] Here, α is the difference between the maximum and minimum values ​​of the feature quantity x. The minimum and maximum values ​​of the feature quantity x may be the value of the feature quantity x in an inactive state. For example, the minimum and maximum values ​​of the feature quantity x may be the resting heart rate HR ｍｉｎ The minimum and maximum values ​​of the feature quantity x can be calculated by using a method that utilizes the result of the context estimation by the behavioral state determination unit 12, or a method that utilizes the sleep heart rate and heart rate variability, but is not limited to these methods. In the method that utilizes the result of the context estimation by the behavioral state determination unit 12, the accuracy of the calculation can be improved by monitoring the user's daily behavioral state. In addition, the maximum heart rate HR ｍａｘ For example, the maximum heart rate HR ｍａｘ You can use statistics such as "220 - Age". ｍａｘ can be estimated from the measured heart rate by utilizing the estimation result of the context. ｍａｘ The method for estimating is not limited to these. The difference α between the maximum and minimum values ​​of the feature quantity x is calculated, for example, using the following formula:

[0051]

[0052] As a further variation, the difference α between the maximum and minimum values ​​of the feature quantity x can be corrected by constants β and γ. Specifically, the values ​​of the constants β and γ can be varied according to user attributes. The user attributes may include, for example, at least one of the user's exercise attributes (attributes of whether or not the user exercises regularly), the user's age, the user's gender, and the time period during which the user wears the biosensor device (the time period during which the biosignal is measured). The variation of the values ​​of the constants β and γ can also be regressively determined based on statistical measurement data according to the user attributes so as to fall into a similar distribution regardless of individual differences. The correction gain Ga of the feature quantity x is calculated by the variation range σ(^) of the user's feature quantity x calculated by the above formula (1) and the value (σ ｍｏｄｅｌ ) and is calculated by the following formula (2).

[0053]

[0054] On the other hand, if it is determined that the amount of data of the measured biological signals stored in the data storage unit 14A is above a certain level (step S400; Y), the emotional response range correction gain calculation unit 14 estimates the fluctuation range σ(^) of the feature x corresponding to the emotional change from the short-term measured biological signal data, and calculates the correction gain Ga.

[0055] 5 is an explanatory diagram showing an example of the correction gain calculation process by the emotional response range correction gain calculation unit 14 when the amount of accumulated data of the actual measured biological signal is equal to or greater than a certain level. In FIG. 5, the horizontal axis represents the time when the biological signal was measured, and the vertical axis represents heart rate variability (Root Mean Square of Successive Differences (RMSSD)). In FIG. 5, a larger value of heart rate variability indicates a more relaxed state.

[0056] The processing flow of the emotional response range correction gain calculation unit 14 when the amount of accumulated measured biological signal data is above a certain level is characterized by buffering data for inactive states and data with guaranteed signal quality, and correcting the feature range for each individual. Since motor physiological responses are mixed in active states, buffering data for inactive states allows for the extraction of mental physiological responses. Specifically, taking into account the characteristics of circadian rhythms, feature x data is buffered for each time period over multiple days (step S402). Furthermore, the fluctuation range σ(^) of the user's feature x is calculated from the maximum and minimum values ​​of the buffered feature x data for each time period over multiple days (step S403). At this time, the signal quality determined by the signal quality determination unit 13 and the activity state determined by the behavioral state determination unit 12 are also taken into account, and a correction gain Ga with reliability is finally calculated (step S404). The fluctuation range σ(^) of the user's feature x is calculated using the following equation (3):

[0057]

[0058] In the above formula (3), A Ｔｋ represents a set of feature quantities x buffered in the time period Tk. ｋ is the reliability of the estimation according to each time period Tk. ｋ may be adjusted and determined based on the approximation error of the derivative function f in the relevant time period, the actual measurement time, the signal quality, etc., but the reliability w ｋ The method for determining the reliability w is not limited to these methods. ｋ can also be determined based on attribute information such as a person's gender, age, and genes.

[0059] The correction gain Ga is calculated by the above formula (2). When this correction gain Ga is used in an actual system, it is considered that the value may not necessarily be stable depending on the user's sensor usage and wearing conditions. Therefore, a function may be provided to update and stabilize the value of the correction gain Ga according to the degree of accumulation of data on the value of the correction gain Ga. For example, data on the value of the correction gain Ga for one week may be buffered, and the reliability w of the value of the correction gain Ga estimated each day may be calculated. ｋ The estimated reliability for each day may be calculated by the sum of the above, and a representative value for each day may be calculated by weighted addition. However, the method for calculating the representative value is not limited to this. The representative value may also be calculated using a linear or non-linear regression curve.

[0060] [1.3 Modification] Fig. 6 shows a schematic configuration example of an information processing device according to a modification. Fig. 7 shows a modification of the correction gain calculation process by the emotional response range correction gain calculation unit 14 when the amount of accumulated measured biological signal data is equal to or greater than a certain level. In Fig. 7, the horizontal axis represents the time when the biological signal was measured, and the vertical axis represents heart rate variability (RMSSD). In Fig. 7, a larger value of heart rate variability indicates a more relaxed state.

[0061] In the information processing device according to the modification, the emotion estimation unit 10 further includes a baseline feature calculation unit 17. The baseline feature calculation unit 17 calculates a baseline feature x ｂａｓｅｌｉｎｅ The baseline feature calculation unit 17 receives as input the feature x data from the feature extraction unit 11, the behavioral state data from the behavioral state determination unit 12, and the signal quality data from the signal quality determination unit 13.

[0062] In this modification, the baseline feature x calculated by the baseline feature calculation unit 17 is used in the emotional response range correction gain calculation unit 14. ｂａｓｅｌｉｎｅ This is because the reference state of a person is calculated using the baseline feature x ｂａｓｅｌｉｎｅ This means that the baseline feature x is used as a baseline feature value and the range of the physiological response is corrected according to the level of arousal. ｂａｓｅｌｉｎｅThe fluctuation range σ(^) of the feature amount x from is calculated by the following formula (4).

[0063]

[0064] According to this modification, it is possible to control the reference state at rest to address the issue of individual differences in physiological responses, and to provide highly accurate emotion estimation technology and services tailored to the user.

[0065] [1.4 Effects] As described above, according to an information processing device according to one embodiment, a physiological response that contributes to a user's emotion is extracted as a feature quantity x, and a correction gain Ga for correcting individual differences in the physiological response that accompanies a change in the user's emotion is calculated from the extracted feature quantity x. Then, the extracted feature quantity x is normalized by a predetermined coefficient (σ ｍｏｄｅｌ ) and the calculated correction gain Ga. This makes it possible to perform highly accurate emotion estimation that is less affected by individual differences in physiological responses.

[0066] In an information processing device according to one embodiment, the emotional response range correction gain calculation unit 14 switches the calculation method for the correction gain Ga depending on the amount of accumulated data of the user's actual measured biological signals, and performs step-by-step correction of the correction gain Ga. When the amount of accumulated data of the user's actual measured biological signals is less than a certain level, the emotional response range correction gain calculation unit 14 calculates the fluctuation range of the feature quantity x corresponding to the fluctuation range of the user's physiological response based on the user's user attributes. On the other hand, when the amount of accumulated data of the user's actual measured biological signals is equal to or greater than a certain level, the emotional response range correction gain calculation unit 14 utilizes the characteristics of circadian rhythms to estimate the fluctuation range σ(^) of the feature quantity x corresponding to emotional changes from short-term measured data. This allows users to enjoy application services based on highly accurate emotion estimation with individual differences corrected, regardless of the length of time they have worn the sensor device 30. Even if a user has only recently worn the sensor device 30, they can enjoy application services based on highly accurate emotion estimation with individual differences optimally corrected using statistical values ​​of user attributes. Furthermore, if the user has a long history of wearing the sensor device 30, the user can enjoy the service of an application that provides more accurate emotion estimation in which individual differences are optimally corrected using the actual data characteristics of the individual.

[0067] The effects described in this specification are merely examples and are not limiting, and other effects may also be achieved. The same applies to the effects of other embodiments described below.

[0068] 2. Other Embodiments The technology according to the present disclosure is not limited to the description of the above embodiment, and various modifications are possible.

[0069] For example, the present technology can be configured as follows. According to the present technology configured as follows, physiological responses that contribute to a user's emotions are extracted as features, a correction gain for correcting individual differences in physiological responses associated with changes in the user's emotions is calculated from the extracted features, and the extracted features are normalized based on a predetermined normalization coefficient and the calculated correction gain. This enables highly accurate emotion estimation that is less affected by individual differences in physiological responses.

[0070] (1) An information processing device comprising: a feature extraction unit that extracts, from a biosignal of a user, a physiological response that contributes to the emotion of the user as a feature; a correction gain calculation unit that calculates a correction gain from the extracted feature to correct individual differences in the physiological response due to emotional changes of the user; and a normalization unit that normalizes the extracted feature based on a predetermined normalization coefficient and the calculated correction gain. (2) The information processing device according to (1), further comprising: an emotional state determination unit that determines the emotional state of the user based on the normalized feature. (3) The information processing device according to (1) or (2), wherein the correction gain calculation unit switches a calculation method of the correction gain depending on the amount of accumulated data of the measured biosignal of the user. (4) The information processing device according to (3) above, wherein, when the accumulated data amount of the measured biological signal of the user is less than a certain level, the correction gain calculation unit calculates a fluctuation range of the feature amount corresponding to a fluctuation range of the physiological response of the user based on user attributes of the user, and calculates the correction gain based on the calculated fluctuation range of the feature amount. (5) The information processing device according to (4) above, wherein the user attributes include an exercise attribute of the user. (6) The information processing device according to (4) or (5) above, wherein the user attributes include at least one of age and gender of the user. (7) The information processing device according to any one of (4) to (6) above, wherein the user attributes include a time period when the biological signal was measured. (8) The information processing device according to any one of (1) to (7) above, wherein the correction gain calculation unit calculates a fluctuation range of the feature amount based on the feature amount when the user is in an inactive state. (9) The information processing device according to any one of (3) to (8), wherein, when the accumulated amount of data of the measured biological signals of the user is equal to or greater than a certain level, the correction gain calculation unit calculates a fluctuation range of the feature amount corresponding to a fluctuation range of the physiological response based on data of the feature amount over a plurality of days accumulated at least during a time period in which emotion estimation of the user is performed, and calculates the correction gain based on the calculated fluctuation range of the feature amount.(10) The information processing device according to any one of (9) to (10), wherein the correction gain calculation unit calculates a representative value of the correction gain for each day based on the feature data for each time period when there are multiple time periods in which the accumulated data amount of the measured biological signal of the user is equal to or greater than a certain level. (11) The information processing device according to (10), wherein the correction gain calculation unit calculates a representative value of the correction gain for each day based on a weighted average of the fluctuation range of the feature for each time period based on the reliability of the fluctuation range of the feature for each time period. (12) The information processing device according to any one of (9) to (10), wherein the correction gain calculation unit updates the value of the correction gain for each day depending on the degree of accumulation of data of the correction gain value for each day. (13) The information processing device according to any one of (1) to (12), further comprising: a baseline feature amount calculation unit that calculates a baseline feature amount indicating a reference state of the physiological response, wherein the correction gain calculation unit calculates a variation range of the feature amount from the baseline feature amount that corresponds to a variation range of the physiological response of the user, and calculates the correction gain based on the calculated variation range of the feature amount from the baseline feature amount. (14) The information processing device according to any one of (1) to (13), further comprising: a signal quality determination unit that determines a signal quality of the biological signal, wherein the correction gain calculation unit calculates the correction gain taking into account the signal quality of the biological signal. (15) The information processing device according to any one of (1) to (14), further comprising: a behavioral state determination unit that determines a behavioral state of the user based on physical information, wherein the correction gain calculation unit calculates the correction gain taking into account the behavioral state of the user. (16) An information processing method including: extracting, from a biosignal of a user, a physiological response that contributes to the emotion of the user as a feature; calculating, from the extracted feature, a correction gain for correcting individual differences in the physiological response due to changes in the emotion of the user; and normalizing the extracted feature based on a predetermined normalization coefficient and the calculated correction gain.

[0071] This application claims priority based on Japanese Patent Application No. 2024-082224, filed on May 20, 2024, in the Japan Patent Office, the entire contents of which are incorporated herein by reference.

[0072] Those skilled in the art will recognize that various modifications, combinations, subcombinations, and variations may occur depending on design requirements and other factors, and are intended to be within the scope of the appended claims and their equivalents.

Claims

1. An information processing device comprising: a feature extraction unit that extracts physiological responses that contribute to the emotions of a user from a biosignal of the user as features; a correction gain calculation unit that calculates a correction gain from the extracted features to correct individual differences in the physiological responses that accompany changes in the emotions of the user; and a normalization unit that normalizes the extracted features based on a predetermined normalization coefficient and the calculated correction gain.

2. The information processing device according to claim 1, further comprising an emotional state determination unit that determines the emotional state of the user based on the normalized feature amount.

3. The information processing device according to claim 1, wherein the correction gain calculation section switches the calculation method of the correction gain depending on the amount of stored data of the user's actual measured biological signal.

4. The information processing device described in claim 3, wherein, when the amount of accumulated data of the measured biological signals of the user does not reach a certain level, the correction gain calculation unit calculates a fluctuation range of the feature corresponding to the fluctuation range of the physiological response of the user based on the user attributes of the user, and calculates the correction gain based on the calculated fluctuation range of the feature.

5. The information processing device according to claim 4, wherein the user attributes include exercise attributes of the user.

6. The information processing device according to claim 4, wherein the user attributes include at least one of the user's age and gender.

7. The information processing device according to claim 4, wherein the user attributes include a time period when the biological signal was measured.

8. The information processing device according to claim 4, wherein the correction gain calculation unit calculates a variation range of the feature amount based on the feature amount when the user is in an inactive state.

9. The information processing device of claim 3, wherein when the amount of accumulated data of the measured biological signals of the user is equal to or greater than a certain level, the correction gain calculation unit calculates a fluctuation range of the feature amount corresponding to a fluctuation range of the physiological response based on data of the feature amount accumulated over multiple days during at least the time period in which the emotion of the user is estimated, and calculates the correction gain based on the calculated fluctuation range of the feature amount.

10. The information processing device described in claim 9, wherein, when there are multiple time periods during which the amount of accumulated data on the user's actual measured biological signals is above a certain level, the correction gain calculation unit calculates a representative value of the correction gain for each day based on the feature data for each time period.

11. The information processing device according to claim 10, wherein the correction gain calculation unit calculates a representative value of the correction gain for each day based on a weighted average of the fluctuation range of the feature amount for each time period based on the reliability of the fluctuation range of the feature amount for each time period.

12. The information processing device according to claim 9, wherein the correction gain calculation unit updates the correction gain value for each day depending on the degree of accumulation of data on the correction gain value for each day.

13. The information processing device according to claim 1, further comprising a baseline feature calculation unit that calculates a baseline feature indicating a reference state of the physiological response, wherein the correction gain calculation unit calculates a variation range of the feature from the baseline feature that corresponds to a variation range of the physiological response of the user, and calculates the correction gain based on the calculated variation range of the feature from the baseline feature.

14. The information processing device according to claim 1, further comprising a signal quality determination unit that determines the signal quality of the biological signal, wherein the correction gain calculation unit calculates the correction gain taking into account the signal quality of the biological signal.

15. An information processing device as described in claim 1, further comprising a behavioral state determination unit that determines the behavioral state of the user based on physical information, wherein the correction gain calculation unit calculates the correction gain taking into account the behavioral state of the user.

16. An information processing method comprising: extracting, from a user's biometric signal, a physiological response that contributes to the user's emotion as a feature; calculating, from the extracted feature, a correction gain for correcting individual differences in the physiological response due to changes in the user's emotion; and normalizing the extracted feature based on a predetermined normalization coefficient and the calculated correction gain.

Images