Gesture detection in embedded applications
Disclosed herein are systems, devices, and processes for gesture detection. A method includes capturing a series of images. The method includes generating motion isolation information based on the series of images. The method includes generating a composite image based on the motion isolation information. The method includes determining a gesture based on the composite image. The processes described herein may include the use of convolutional neural networks on a series of time-related images to perform gesture detection on embedded systems or devices.
1. A method, comprising:
capturing a series of images;
generating motion isolation information based on the series of images, wherein generating motion isolation information includes flattening pixel values in each image of the series of images based on a predetermined threshold pixel value;
generating a composite image based on the motion isolation information, wherein generating the composite image includes applying a color gradient, wherein applying the color gradient includes determining a series of evenly spaced color intervals based on the number of images in the series of images that are to be merged to generate the composite image, wherein applying the color gradient includes applying the darkest remaining color value of the series of evenly spaced color intervals to the oldest remaining image of the series of images that are to be merged to generate the composite image; and
determining a gesture based on the composite image.
2. The method of claim 1 , wherein generating the composite image includes merging the motion isolation information.
3. The method of claim 2 , wherein the motion isolation information is a second series of images, and
wherein merging the motion isolation information includes combining portions of more than one of the second series of images to form the composite image.
4. The method of claim 3 , wherein the combining portions of more than one of the second series of images includes applying at least one set of overlapping portions of the more than one of the second series of images.
5. The method of claim 4 , wherein the at least one set from the at least one set of overlapping portions includes a first portion to which a first color from the color gradient is applied, and a second portion to which a second color from the color gradient is applied.
6. The method of claim 1 , wherein determining the gesture based on the composite image includes applying the composite image as input to an artificial neural network.
7. The method of claim 6 , wherein the artificial neural network is a convolutional neural network.
8. The method of claim 7 , wherein the convolutional neural network is not a recurrent neural network.
9. The method of claim 7 , wherein the determining the gesture based on the composite image is performed by the processor of an embedded system.