Ovládaní počítače nevidomou osobou

Abstract

This bachelor thesis deals with the analysis of speech signal in the area of time for purposes of controlling the computer via vocal commands. Resultant program solution contains in the first place calibration of the microphone due to possible ambient disturbances. Consequently the user is invited to train (record) voice commands assigned to individual program functions. These command samples are divided into segments (each 25 ms long) and consecutively analysed by means of 3 parameters – zero crossing, short-term intensity and ACF. Then they are saved to the dictionary through the serialization mechanism. Application trained this way is ready to catch the voice acoustic speech signals and compare them with samples in the dictionary using the DTW algorithm. If the results of comparison are positive, the program will carry out the specific actions assigned to particular voice command