Sorting of autocorrelation portrais to accelerate speech commands recognition

We investigate a method to accelerate the recognition algorithm of the speech signal using the phase prior exclusion from consideration deliberately mismatched autocorrelation portraits of voice commands. At the stage of pre-processing for each portrait created a small copy of which is used for a quick decision on the similarity with the reference team delivered. In the case of a positive decision, ​​a comparison of the full-size image is performed. Application of the algorithm has reduced the maximum time command recognition twice, while maintaining the efficiency of recognition.