语音增强的长大(国外英文资料).docVIP

语音增强的发展(国外英文资料) Voice enhancement At the beginning of the 20th century and interim progress mainly in the field of speech enhancement in the acoustic field in the study of speech intelligibility, and the main progress in recent decades is from the digital signal processing and pattern recognition. In the sixties and seventies, the research on the intelligibility of speech fell into silence. Another source dynamic signal processing of speech enhancement research begins to show its power. This source also originated from world war ii, which was the key to automatic mechanical control theory and electronic computer. Smooth wiener is proposed under the condition of wiener filtering theory, and after the war the rapid development of space technology, such rocket and satellite, led to the state variables in recursive filtering theory of kalman filter. But because the voice signals are unsteady and cannot be used in state equations, the filter theory was not used for speech enhancement. In mid and late seventy s, the progress of semiconductor development led to a general computer, digital signal processing technology developed so quickly, its speech processing real turning point. The rapid Fourier Transform (fft-fast Fourier Transform) and the development of linear systems theory propose a model of speech sound, as shown in figure 1-3. Figure 1-3 speech production model The Av and Au respectively are the excitation amplitude of the voiceless and the sound, the transmission function of the channel model V (z) is approximately denoted by the full-pole model (ar-auto Regressive). The N is the number of models, the coefficients on each of the first poles, and z is the independent variable of the z transform domain. This model can be equivalent to the Linear Predictive Code (lpc-linear Predictive Code) model, which is the rapid recursive method of the output signal solving model coefficients. On the basis of the above model, with the aid of the computers powerful computing abi