Method for extracting feature vectors for speech recognition

Abstract

Disclosed is a method for speech recognition which achieves a high recognition rate. The method includes extracting a parameter from an input signal that represents a characterization of the input signal as a voiced or unvoiced sound, extracting at least one feature vector corresponding to an overall spectrum shape of a voice from an input signal, and using the extracted parameter and extracted feature vectors in a training phase and in a recognition phase to recognize speech.

Description

technical field [0001] The invention relates to speech recognition, in particular to a method for extracting feature vectors to achieve high speech recognition rate. Background technique [0002] In the neighborhood of speech recognition, the two speech recognition methods that are mainly used are Hidden Markov Model (HMM) and Dynamic Time Warping (DTW). [0003] In the HMM-based speech recognition method, HMM parameters are acquired in the training stage and stored in a speech database, and a Markov processor searches for a model with the highest recognition rate using a maximum likelihood (ML) method. Feature vectors necessary for speech recognition are extracted, and training and speech recognition are performed using the extracted feature vectors. [0004] In the training phase, the HMM parameters are usually obtained using the Expectation Maximum (EM) algorithm or the Baum-Welch re-estimation algorithm. The Viterbi algorithm is usually used in the speech recognition s...

AI Technical Summary

Problems solved by technology

However, a limitation of these methods is that they do not take into account the difference between the voiced and non-voiced sounds that make up the speech

Images