Research On Signal Processing Of Individual Characteristics Based On Speech Recognition

In order to make the machine better understand users' requirements and achieve the artificial intelligence,it's necessary to use voice input instead of traditional mouse and keyboard input mode.The speech signal contains a large number of complex information in the two-dimensional signal.It also has time variant and non standard.The recognition and processing of the speech signal should start with the personality characteristic signal.In this paper,we first study human voice and auditory system.The principle of human voice is analyzed,and the digitized model and characteristics of voice production are sorted out accordingly.And we study the auditory system of the human ear which is the evaluation standard for speech recognition.In order to extract the individual characteristic signals in speech,the analysis method of speech signal in time domain,frequency domain and inversion frequency domain is studied.The whole pole model is established and solved by autocorrelation method,and obtain the linear prediction coefficient which is very important for the extraction of personality characteristics.Next,in order to make the extracted personal characteristic signal accurate and clear,we first preprocess the speech signal.We design experiment and compare the effects of various digital filters.Choose the best filters from different speech signals.There are three kinds of speech signals in personality.According to the actual situation,we focus on the extraction and processing of pitch and formant parameters.In the extraction process,the fundamental frequency is extracted by cepstrum,so that the influence of formant on the extraction process is eliminated,and the results are smoothed and optimized.The formant parameters are estimated by linear prediction method and solved by two times interpolation method.In the end,linear predictive value is used instead of single frame resampling to change the speed of speech signal,and a good interpolating line spectrum is introduced to carry out the speech transmission experiment.The personal characteristic signal is processed through Matlab simulation,and achieves better speed change and sandhi effect.