Emotion and action tracking and tracking for video conferencing
The present disclosure provides systems and methods for extraction of nonverbal communication data from video. A system can include a computing device comprising a processor and a camera. The system can retrieve, from a camera of the computing device, a video stream of a user. The system can select a plurality of individual frames of the video stream. For each of the plurality of individual frames of the video stream, the system can extract a plurality of features and identify, from the extracted plurality of features, a pose of the user. The system can classify, via a neural network from the identified poses of the user for the plurality of individual frames of the video stream, the video stream as showing one of a predetermined plurality of states.
1. A method for extraction of nonverbal communication data from video, comprising:
retrieving, by a processor of a computing device from a camera of the computing device, a video stream of a user;
selecting, by the processor, a plurality of individual frames of the video stream;
for each of the plurality of individual frames of the video stream:
extracting, by the processor, a plurality of features,
identifying, from the extracted plurality of features for a frame of the plurality of individual frames, a plurality of points corresponding to physical features of the user at identified positions within the frame, and
identifying, by the processor from the extracted plurality of features, a pose of the user; and
classifying, by the processor via a neural network from the identified poses of the user for the plurality of individual frames of the video stream, the video stream as showing one of a predetermined plurality of states by providing an array comprising identified positions of points corresponding to a physical feature of the user for each of the plurality of individual frames to the neural network, and identifying a highest scoring state of the plurality of states from output values of the neural network.
2. The method of claim 1 , wherein retrieving the video stream of the user further comprises retrieving the video stream from one of a camera connected to the computing device and a camera connected to a remote computing device.
3. The method of claim 1 , wherein selecting the plurality of individual frames of the video stream further comprises downsampling the video stream at a predetermined rate, and selecting the plurality of individual frames from consecutive frames of the downsampled video stream.
4. The method of claim 1 , wherein extracting the plurality of features further comprises processing each frame of the plurality of individual frames with a convolutional neural network.
5. The method of claim 1 , wherein extracting the plurality of features further comprises performing an edge detection and dimensional reduction on each frame of the plurality of individual frames.
6. The method of claim 1 , wherein classifying the video stream as showing one of the predetermined plurality of states further comprises:
providing a plurality of arrays comprising identified positions of points corresponding to different physical features of the user for each of the plurality of individual frames to a corresponding plurality of neural networks; and
identifying a highest scoring state of the plurality of states from aggregated output values of the plurality of neural networks.
7. The method of claim 1 , further comprising providing an identification of the classified state of the video stream to a remote computing device.
8. The method of claim 7 , wherein the video stream is not provided to the remote computing device.
9. A system for extraction of nonverbal communication data from video, comprising:
a computing device comprising a processor and a camera;
wherein the camera is configured to capture a video stream of a user; and
wherein the processor is configured to:
select a plurality of individual frames of the video stream,
for each of the plurality of individual frames of the video stream:
extract a plurality of features;
identify, from the extracted plurality of features for a frame of the plurality of individual frames, a plurality of points corresponding to physical features of the user at identified positions within the frame; and
identify, from the extracted plurality of features, a pose of the user, and
classify, via a neural network from the identified poses of the user for the plurality of individual frames of the video stream, the video stream as showing one of a predetermined plurality of states by providing an array comprising identified positions of points corresponding to a physical feature of the user for each of the plurality of individual frames to the neural network, and identifying a highest scoring state of the plurality of states from output values of the neural network.
10. The system of claim 9 , wherein the processor is further configured to downsample the video stream at a predetermined rate, and select the plurality of individual frames from consecutive frames of the downsampled video stream.
11. The system of claim 9 , wherein the processor is further configured to process each frame of the plurality of individual frames with a convolutional neural network.
12. The system of claim 9 , wherein the processor is further configured to perform an edge detection and dimensional reduction on each frame of the plurality of individual frames.
13. The system of claim 9 , wherein the processor is further configured to:
provide a plurality of arrays comprising identified positions of points corresponding to different physical features of the user for each of the plurality of individual frames to a corresponding plurality of neural networks; and
identify a highest scoring state of the plurality of states from aggregated output values of the plurality of neural networks.
14. The system of claim 9 , wherein the processor is further configured to provide an identification of the classified state of the video stream to a remote computing device.
15. The system of claim 14 , wherein the video stream is not provided to the remote computing device.
16. A non-transitory computer-readable medium comprising instructions that, when executed by a processor of a computing device, cause the processor to perform operations comprising:
retrieving, from a camera of the computing device, a video stream of a user;
selecting a plurality of individual frames of the video stream;
for each of the plurality of individual frames of the video stream:
extracting a plurality of features,
identifying, from the extracted plurality of features for a frame of the plurality of individual frames, a plurality of points corresponding to physical features of the user at identified positions within the frame; and
identifying, from the extracted plurality of features, a pose of the user, and
classifying, via a neural network from the identified poses of the user for the plurality of individual frames of the video stream, the video stream as showing one of a predetermined plurality of states by providing an array comprising identified positions of points corresponding to a physical feature of the user for each of the plurality of individual frames to the neural network, and identifying a highest scoring state of the plurality of states from output values of the neural network.