Implementation of Voice Recognition Via CNN and LSTM

Abstract

The voice recognition system uses CNN a lot. This is because CNN has the optimized ability to recognize and classify targets. CNN, however, has a problem that the bigger the object to be recognized, the more expensive the computational costs are. In this paper, we are going to solve these problems through MFCC feature extraction and model roll combining CNN and LSTM to present the possibility of performing voice recognition even through low-cost devices.