Embedded voice recognition system based on improved MFCC algorithm and method thereof

The invention relates to an embedded voice recognition system based on an improved MFCC algorithm and a method thereof. The method comprises the following steps: entering an identification mode, collecting a user voice signal and performing preprocessing; performing endpoint detection on the voice signal to determine a starting point and an ending point of the voice signal; and performing featureextraction on the voice signal to obtain a 12-dimensional mixed feature parameter, which can be all features of the voice signal input by the user; matching the obtained feature parameter with the feature templates pre-stored in a main control system, if matched, obtaining the final recognition result; otherwise, re-executing a recognition mode; wirelessly controlling the user-specified device bya Bluetooth module; performing selection to enter a training mode; selecting the storage points of the training templates by the user, if the storage point of each template already has a template, selecting one of the storage