Research On The Pick-up Algorithm Ofphonetic Enigenvalue Based On The Wavelet Transform

The computer is playing an increasingly important role in human life and human expect to communicate with computer in the direct and rapid way. We hope the computer understand human language. It can instead of relying on the keyboard and mouse input solely. The technique of speech recognition can satisfy the desire of mankind. Phonetic recognition is using speech signal as the research object. It is the process of human sound converts into the text or instructions, and its ultimate aim is to make the natural voice become the way of communication between human and the computers. In many fields of human life, the speech recognition technology has been widely applied. It provides more convenience for human life, so it has broad application prospects.Characteristic Parameter Extract is the top priority of the speech recognition. It is a process to extract the information from the abundant speech signal which is helpful to recognize the useful information and to analyze the speech signal to wipe out the useless information which is useless to speech recognition. Pitch period is a very important characteristic parameter of the speech characteristic value which play a very important role in the speech coding, speech synthesis, speaker recognition and speech recognition, so checking the importance of the pitch period is very important. The article use the Wavelet transform to do the noise suppression for original speech signal and to do the extraction of the EACF pitch period. In the condition of the Low SNR, the traditional autocorrelation function may easily happen to cycle and times cycle mistakes. To solve these problems, inspired of the method of harmonic product spectrum, the article propose the detecting cycle algorithm of the Harmonic spectrum pitch deposition which based on the EACF. The experiment has proved that this method can mark out the unvoiced or voiced sound accurately and detect the Pitch curve smooth and describe the changes of the pitch period accurately which can prove the reliability of the pitch period detection.