A Wavelet Transform Coding For Voice's DCT

Owing to the rapid development of the computer and network technology, multimedia plays a significant role in it, voice is one of the very basic and natural media for human being to communication and information changing. The technology on speech signal processing has a tremendous potential and advantages, and can be used in many field, such as communications area, computer network, industrial control, education and entertainment. Digital speech signal has many virtues, such as easy to saving and transmission, to store with less memory, and good robust, high secrete and low cost, so it will have more and more field to be used in today's digitalization and enormous information era.In applications, we always consider these two problems , one is the better quality of reproduced voice with small storage requirement and less data rate in transmission, another is very simple in processing . The key for these problems is to used quit good and efficient voice coding algorithm. This is a very attractive work for me to do.After studying the characters of speech signal, a new speech coding algorithm is presented, which is called wavelet coding for voice discrete cosine transform. The discrete cosine transform of voice signal that is a transformation in frequency domain shows energy distribution. If we can approximate the energy distribution curve in frequency domain according to the ear's frequency character, therefore we can get a very good quality reproduced voice in subject opinion. The wavelet transform of signal is the modern signal processing method coming from the traditional frequency analyzing of signal, especially in STFT(Short Time Fourier transform ) . During many years' theoretical and applied study, it is superior to other tools on processing time-varying signals. In this paper, we do some studies on speech compressing with Wavelet transformation.In order to better and easier control reproduced voice quality, we can use different wavelets and analysis method( wavelet transform or wavelet packet transform), also can choice different analysis level.The study platform is based on MATLAB 6.2, encoding/decoding programs are also presented by using MATLAB 's program. We use wavelet packet transform of wavelet 'sym2' with 2 analysis level (encoding 4 coefficients ). The male voiceis used as original speech signal digitalized in u-law 8k Hz. By encoding and decoding ,a very high quality voice seemly original one is produced in 16k bps data rate.