Consultation system and consultation method

Abstract

This invention discloses a consultation system and method. The consultation system continuously captures the user's physiological signals to establish a physiological baseline and calculates the amount of physiological change. Then, it uses an emotion mapping model to deduce a current emotional feature vector. During the consultation process, if the difference between the current emotional feature vector and the expected emotion of the current consultation dialogue script exceeds a first preset threshold, the system pauses the current script. Next, the system calculates the difference between the current emotional feature vector and other scripts in the database. If the minimum difference is less than a second preset threshold, the system automatically loads and switches to the corresponding consultation dialogue script. If the minimum difference still does not meet the condition, the system outputs an interruption signal and terminates the program. Therefore, the system can adapt or interrupt the consultation content based on the user's actual emotional feedback in real time, improving the accuracy and safety of the consultation process.

Claims

1. A consultation system, comprising: A physiological signal sensing device that captures multiple physiological signals from a user; A database stores multiple counseling dialogue scripts and an emotion mapping model, each of the multiple counseling dialogue scripts having a predicted emotion feature vector; and a processing unit coupled to the physiological signal sensing device and the database, the processing unit performing the following operations: loading one of the counseling dialogue scripts from the database as a current counseling dialogue script; continuously receiving the multiple physiological signals captured by the physiological signal sensing device within a preset time period, calculating an average value of each of the multiple physiological signals within the preset time period, generating multiple physiological baselines; after generating the multiple physiological baselines, repeatedly performing the following operations: calculating multiple physiological changes in the multiple physiological signals continuously captured by the physiological signal sensing device and the corresponding multiple physiological baselines; combining the multiple physiological changes into a physiological feature vector. The physiological feature vector is input into the emotion mapping model, which outputs a current emotion feature vector. When a first difference between the current emotion feature vector and the expected emotion feature vector of the current counseling dialogue script is greater than a first preset threshold, execution of the current counseling dialogue script is paused, and multiple second differences between the current emotion feature vector and the expected emotion feature vectors of the multiple counseling dialogue scripts are calculated. When the smallest difference among the multiple second differences is less than a second preset threshold, one of the counseling dialogue scripts corresponding to the smallest difference is loaded from the database to replace the current counseling dialogue script. When the smallest difference is not less than the second preset threshold, an interrupt signal is output, and the repeated execution of the operation is stopped.

2. The consultation system according to claim 1, wherein, The multiple physiological signals include a brainwave signal, a heart rate variability signal, a skin conductance response signal, and an eye movement signal.

3. The consultation system according to request item 1, wherein, The plurality of physiological signals include an EEG signal, a heart rate variability signal, and a skin conductance response signal. The plurality of physiological changes include a power spectral density change corresponding to the EEG signal, a high-low frequency power ratio change corresponding to the heart rate variability signal, and a phase conductance change corresponding to the skin conductance response signal. The physiological feature vector is composed of the power spectral density change, the high-low frequency power ratio change, and the phase conductance change.

4. The consultation system according to claim 1, wherein, The emotion mapping model includes a convolutional neural network, a recursive neural network, and an attention layer. The processing unit inputs the physiological feature vector into the convolutional neural network to generate a first feature vector, inputs the first feature vector into the recursive neural network to generate a second feature vector, and inputs the second feature vector into the attention layer to output the current emotion feature vector.

5. The consultation system described in Request 1, wherein, Both the current emotion feature vector and the expected emotion feature vector are six-dimensional feature vectors, and the multiple dimensions of the six-dimensional feature vector are respectively an anger value, a fear value, an anxiety value, a sadness value, a depression value and a shame value.

6. A counseling method, executed by a processing unit coupled to a physiological signal sensing device and a database, the database storing multiple counseling dialogue scripts and an emotion mapping model, each of the multiple counseling dialogue scripts having a predicted emotion feature vector, the counseling method comprising the following steps: loading one of the counseling dialogue scripts from the database as a current counseling dialogue script; continuously receiving multiple physiological signals of a user captured by the physiological signal sensing device within a preset time period, calculating an average value of each of the multiple physiological signals within the preset time period, generating multiple physiological baselines; after generating the multiple physiological baselines, repeatedly executing the following steps: calculating multiple physiological changes in the multiple physiological signals continuously captured by the physiological signal sensing device and the corresponding multiple physiological baselines; combining the multiple physiological changes into a physiological feature vector. The physiological feature vector is input into the emotion mapping model, which outputs a current emotion feature vector. When a first difference between the current emotion feature vector and the expected emotion feature vector of the current counseling dialogue script is greater than a first preset threshold, execution of the current counseling dialogue script is paused. Multiple second differences between the current emotion feature vector and the expected emotion feature vectors of the multiple counseling dialogue scripts are calculated. When the smallest difference among the multiple second differences is less than a second preset threshold, one of the counseling dialogue scripts corresponding to the smallest difference is loaded from the database to replace the current counseling dialogue script. When the smallest difference is not less than the second preset threshold, an interrupt signal is output, and the repeated execution of the steps is stopped.

7. The consultation method described in request item 6, wherein, The multiple physiological signals include a brainwave signal, a heart rate variability signal, a skin conductance response signal, and an eye movement signal.

8. The consultation method described in request item 6, wherein, The plurality of physiological signals include an EEG signal, a heart rate variability signal, and a skin conductance response signal. The plurality of physiological changes include a power spectral density change corresponding to the EEG signal, a high-low frequency power ratio change corresponding to the heart rate variability signal, and a phase conductance change corresponding to the skin conductance response signal. The physiological feature vector is composed of the power spectral density change, the high-low frequency power ratio change, and the phase conductance change.

9. The consultation method according to request item 6, wherein, The emotion mapping model includes a convolutional neural network, a recursive neural network, and an attention layer. The step of inputting the physiological feature vector into the emotion mapping model and having the emotion mapping model output a current emotion feature vector includes: inputting the physiological feature vector into the convolutional neural network to generate a first feature vector, inputting the first feature vector into the recursive neural network to generate a second feature vector, and inputting the second feature vector into the attention layer to output the current emotion feature vector.

10. The consultation method according to claim 6, wherein, Both the current emotion feature vector and the expected emotion feature vector are six-dimensional feature vectors, and the multiple dimensions of the six-dimensional feature vector are respectively an anger value, a fear value, an anxiety value, a sadness value, a depression value and a shame value.

Images