Comprehensive Analysis And Application Of Template Matching Algorithm Based On Feature Extraction Of Speech Signal

Speech signal processing is to make the machine can understand the human voice, the most fundamental purpose is to belong to a person or that person’s nature as an important attribute, The principle is that the sampling of the original speech waveform, but human speech includes many elements, there is need, but also we do not need, We should extract that we need to study portions. Because the technology is now mature them, compared to other technologies with unparalleled convenience, low cost and support for remote applications and other advantages, has broad application background in the application. Has been widely used in various aspects, described as pervasive, not only military, but also civilian, such as military aspects related to the military, defense, and military communications. With even more advances in technology, processing of speech signals become our daily life and work in a very large penetration of the main ways you can perform authentication and identification and implementation means that people research and related product development and more attention of researchers in theory.Characteristics of the speech signal extraction is among the key speech signal analysis, which is the most important aspect is the focus of research in related fields and directions. In practical application, the speaker recognition in different environments and even feature extraction can be carried out under the harsh environment, which for us in the experiment that will cause some trouble. Therefore, in the feature extraction process. We can extract the information of the speech signal, not only has better stability, but also has more desirable characteristic parameter, which is the problems we face.In this paper, the following improvements were made and tested. The first experiment is based on hybrid LPCC and MFCC feature extraction algorithm for text-independent. Extracting characteristic parameters of the algorithm combines the advantages and difference coefficients, which improve the recognition rate of the system. The second experiment is based on dynamic time to plan and improve endpoint detection algorithm to achieve an isolated word recognition system. Speakers pronounce "0-9" sampling and analysis of the data, the results show that the improved algorithm compared to traditional DTW algorithm has better recognition rate and speed.