Font Size: a A A

Research Of Man Voice Recognition Technology Based On VxWorks

Posted on:2006-12-26Degree:MasterType:Thesis
Country:ChinaCandidate:Z Y ZhouFull Text:PDF
GTID:2132360155968948Subject:Navigation, guidance and control
Abstract/Summary:PDF Full Text Request
As a very important criterion of the Voyage Data Recorder, the recognition rate and character of real-time of Man Voice Recognition (MVR) are very important to the system for data recording. MVR is the arithmetic to determine whether there is anybody talking by judging some parameters which are collected from the microphone or calculated from the source voice data. After judging these data, it will tell system the result.MVR arithmetic can be concluded by analyzing the source voice data, picking the character parameters and judging in two aspects. One is base on the character of man's throat and voice. The other is based on linguistics. This article gives some practical arithmetic with studying these two aspects.Through analyzing man's throat math model and the character of man's voice in time field, this article gives some arithmetic based on short time zero cross rate, short time average energy and formant. The article also gives the way of realizing MVR by computer.The way that MVR based on linguistics can be divided into two steps. The first step is picking the Linear Predictive Spectrum Parameter or Mel-Frequency Cepstral Coefficients. The second step is matching these parameters with template, and then makes decision. It can decrease the matching speed by simply using Chinese phoneticism matching.In this article, it also gives the idea that can develop the recognition rate by use the method of combination. For some reason one single character may make a error decision that the voice is pronounced of man, but with another character together it may make a better decision.
Keywords/Search Tags:Mel-Frequency Cepstral Coefficients, Man Voice Recognition, Voyage Data Recorder, Short Time Zero Cross Rate, Short Time Average Energy, formant, Linear Predictive Spectrum Parameter
PDF Full Text Request
Related items