Emotion detection using speaker baseline
Described herein is a system for emotion detection in audio data using a speaker's baseline. The baseline may represent a user's speaking style in a neutral emotional state. The system is configured to compare the user's baseline with input audio representing speech from the user to determine a emotion of the user. The system may store multiple baselines for the user, each associated with a different context (e.g., environment, activity, etc.), and select one of the baselines to compare with the input audio based on the contextual situation.
1. A computer-implemented method, comprising:
receiving input audio data;
determining that the input audio data represents audio from a user associated with a user profile;
receiving first context data corresponding to the input audio data;
selecting first reference audio data associated with the user profile, wherein the first reference audio data is selected based on the first context data corresponding to second context data associated with the first reference audio data;
determining first feature data representing acoustic attributes corresponding to the first reference audio data;
determining second feature data representing acoustic attributes corresponding to the input audio data; and
processing the first feature data and the second feature data using a trained model to determine output data representing at least one difference between the input audio data and the first reference audio data.
2. The computer-implemented method of claim 1 , further comprising:
determining the user profile is associated with the first reference audio data and second reference audio data,
wherein the selecting comprises selecting the first reference audio data instead of the second reference audio data.
3. The computer-implemented method of claim 1 , wherein the first context data indicates the user is participating in a first activity.
4. The computer-implemented method of claim 1 , wherein the first context data indicates the user is interacting with another user.
5. The computer-implemented method of claim 1 , wherein the first reference audio data corresponds to baseline speech spoken by the user.
6. The computer-implemented method of claim 1 , wherein the first reference audio data corresponds to a neutral state of the user.
7. The computer-implemented method of claim 1 , wherein the output data further represents a first state of the user corresponding to the audio relative to a second state of the user corresponding to the first reference audio data.
8. The computer-implemented method of claim 1 , wherein determining the first feature data comprises:
processing the first reference audio data using a first encoder to determine the first feature data, the first encoder having at least a first processing layer and a second processing layer.
9. The computer-implemented method of claim 8 , wherein determining the second feature data comprises:
processing the input audio data using a second encoder and data corresponding to the second processing layer to determine the second feature data.
10. A system comprising:
at least one processor; and
at least one memory comprising instructions that, when executed by the at least one processor, cause the system to:
receive input audio data;
determine that the input audio data represents audio from a user associated with a user profile;
receive first context data corresponding to the input audio data;
select first reference audio data associated with the user profile, wherein the first reference audio data is selected based on the first context data corresponding to second context data associated with the first reference audio data;
determine first feature data representing acoustic attributes corresponding to the first reference audio data;
determine second feature data representing acoustic attributes corresponding to the input audio data; and
process the first feature data and the second feature data using a trained model to determine output data representing at least one difference between the input audio data and the first reference audio data.
11. The system of claim 10 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
determine the user profile is associated with the first reference audio data and second reference audio data,
wherein the instructions that cause the system to select the first reference audio data comprise instructions that, when executed by the at least one processor, cause the system to select the first reference audio data instead of the second reference audio data.
12. The system of claim 10 , wherein the first context data indicates the user is participating in a first activity.
13. The system of claim 10 , wherein the first context data indicates the user is interacting with another user.
14. The system of claim 10 , wherein the first reference audio data corresponds to baseline speech spoken by the user.
15. The system of claim 10 , wherein the first reference audio data corresponds to a neutral state of the user.
16. The system of claim 10 , wherein the output data further represents a first state of the user corresponding to the audio relative to a second state of the user corresponding to the first reference audio data.
17. The system of claim 10 , wherein the instructions that cause the system to determine the first feature data comprise instructions that, when executed by the at least one processor, cause the system to:
process the first reference audio data using a first encoder to determine the first feature data, the first encoder having at least a first processing layer and a second processing layer.
18. The system of claim 17 , wherein the instructions that cause the system to determine the second feature data comprise instructions that, when executed by the at least one processor, cause the system to:
process the input audio data using a second encoder and data corresponding to the second processing layer to determine the second feature data.