Research On Speech Enhancement Under Noise Of Music

Recently, Speech Enhancement has become one of the important branches in Speech signal digital processing. Its main functions are reducing noise and enhancing speech clarity, mainly used in voice recognition and voice coding system pre. The way of Speech Enhancement has a lot, whose goal is to enhance the degree of voice clarity and understanding. However, any of Speech Enhancement system is based on different application to make the appropriate choices and trade-offs.Music is a kind of entertainment for our daily life, but when we walk down the street, we can hear large and sometimes small music, especially in KTV, Video stores, even the passers ringtones. The music has produced pollution on our normal communication, and dropped our communication quality. The music noise is an important part of the community noise, noise hazard is obvious. My hoping is that this kind of music pollution can reduce the minimize data from this assignment.Main content of this article are: first of all, everywhere there is noise in the actual environment, understand the characteristics of the noise is an important basis for speech enchancement. Noise is a special audio signal, therefore, the noise can be time-domain and frequency domain analysis. Find out the difference with the speech signal to identify the speech enhancement algorithm. Secondly, music signal as noise analysis, music signal has its own inherent characteristics. To understand the basis of these features, find the adoption of better algorithms to filter out. The final, design is based on the H∞filtering blind adaptive filter from the speech enhancement algorithm. This H∞filter applied to music from the background noise environment, this speech enhancement is only based on the character of speech and background noise music, but the noise which is quasi-stationary process is adaptively extracted from the noisy speech pure voice, without having to predict the statistical properties of music and the noise source parameters. The H∞filter has four main subsystems: H∞pre-filtering subsystem, noise identification subsystem, noise feed-forward solution and the virtual feedback subsystem, through three functions of each subsystem to achieve the best performance. The results show that the basic musical background noise is filtered out, and can get the original voice signal, meanwhile improved voice clarity and comprehension.