The Algorithm Based on Supervised Time Series Segmentation of Researching Dynamic Gesture Recognition

The Algorithm use the changes of pixels in the time domain in image sequences and the correlation between adjacent frames, find the corresponding relationship between the previous frame and the current frame to collect samples, ensure that the whole gesture can be collected without too many redundant frames. Use two-way propagation search mechanism, shortest length constraint l
min
and maximum length constraint l
max
are introduced to reduce the search space. By calculating an estimate ei of a point, determinate of segmentation points for sequence segmentation quantitatively, then calculate the joint probability of each characteristic distribution under each state, category matching based on maximum probability. The experiment uses a large number of samples to train the model and compares it with other algorithms, it shows that the algorithm has achieved good results in dynamic gesture recognition.