Font Size: a A A

Brief Analysis Based On Spectrogram Partial Characteristic Of Chinese Consonants Classification Method

Posted on:2013-03-19Degree:MasterType:Thesis
Country:ChinaCandidate:H Y LiFull Text:PDF
GTID:2248330395971385Subject:Circuits and Systems
Abstract/Summary:PDF Full Text Request
With the rapid development of speech recognition technology,our needs andrequirements of the speech recognition technology have increased year by year,although recent years the speech recognition technology is constantly update,optimization.But the recognition rate of speech recognition has always been on the lowlevel. Relative to the speech recognition, the image recognition technology has a goodbackground, various theories have been more mature.This paper first studied the formerly speech recognition applicationtechnology,and the shortcomings,and then introduced the pronunciation characteristicsof chinese. Propose a new speech recognition method, despite the speech signalwaveform, extract the spectrogram of the speech signal as the processing and identifyobject, this will convert speech recognition to the image recognition. The object of thispaper is the22consonants of Chinese. The22consonants of the Chinese language isdivided into six categories according to the laws,the processing software is MATLAB.Pretreatment process finished the image morphological filtering, because of thedifferent Chinese pronunciation methods and pronunciation parts, extract twoimportant characteristics of the image, geometry characteristic and the resonance peakfeatures. As the recognition characteristics.Then make the the feasibility theoreticalanalysis about the two features applied to the neural network trained and recognize.The analysis shows that the two features can be fully applied to the spectrogram imagerecognition, so as to achieve the purpose of speech recognition.The thesis in spectrum diagram language recognition direction of the research hasjust started, confirm the spectrum diagram language used to identify the feasibility ofspeech recognition, hope in the future research spectrum diagram language recognitioncan get better application, to realize a more accurate speech recognition.
Keywords/Search Tags:Speech recognition, Spectrogram, Resonance peak, Mathematicalmorphology
PDF Full Text Request
Related items