Study On The Key Technology Of Multimedia Language Leaning System

In order to reduce the lab-background noise in noisy speech to minimum, this paper studies Speech enhancement to some extent.Firstly, analyze the primary Dynamic noise reduction (DNR) model, and present an improvement project. Then based on the acoustics' and the phonetics' pertinent conclusion, the Wavelet-wiener filtering (WWF) algorithm is put forward according to the wavelet analysis and Wiener filtering.Secondly, Pre-emphasis is used to improve the high boost in WWF algorithm. In the same way, using an effective frame regrouping precept improve frame overlapping and regrouping, designing an new SH window to window function's design, selecting coif wavelet as the analysis wavelet basis to wavelet basis's selection, discarding the unvoiced frames by threshold-wavelet transform to unvoiced/voiced frames' distinguish, and solving the wiener gain by the fake cepstral subtraction to Wiener gain's getting as well.Thirdly, complete WWF model in Matlab. According to the two means, signal-to-noise ratio (SNR) and I-Satio spectrum distortion measure, WWF model is evaluated in the ways of the single frame and the whole signal. And the result is analyzed in brief.Lastly, four effective and feasible ways are planed to improve WWF model. And to compensate the enhanced signal's energy loss and spectrum distortion, several reasonable schemes of Speech compensation are touched upon in the paper.Aiming at speech enhancement's design specification of the primary system, WWF model is established in the paper, which combines the wavelet analysis and wiener filtering. And the evaluation indicates WWF model's effectiveness. In addition, the improvement projects and the thoughts on Speech compensation are mentioned here, which can be good bases to consummate system's functions ulteriorly.