Speech Enhancement Technology With Application To Digital Communications

Digital speech communication is becoming more and more important in our society with the digital information time's coming. The purpose of speech enhancement is to improve the quality of the speech intension, length, tone and pureness by transforming and processing signals in time or frequency domain, while the meaning and explicitness of speech should not be destroyed. The performance of the speech process system and the quality of the digital speech communication can also be improved in the disturbed environment by enhancement implement.An automatic level control system for speech enhancement is discussed in this thesis. This system is necessary to complete the detection of speech and audio, and then an automatic gain control method should be used. It should also be consistent with the protocols and standards. As running in a digital communication network, the automatic level control system should not be effect in the non-speech data transmission with the help of some proper detection algorithms, and should be able to recognize the characteristics of the speech signal before the level has been adjusted properly. This thesis aims to construct a flexible and stable speech enhancement system in accord with ITU-T G.I69. For this purpose, the constructed system must be adapted to the requirements of the digital communication network and preferable performances can be achieved through the experiments. The research is divided into three parts:First, we are concerned about the voice activity detection algorithms based on the spectrum characteristics and the single frequency signal rapid detection. The voice activity detection, which distinguishes signals between speech and non-speech statement, is the basic section of many speech signal process system. Our voice activity detection method is simple to be implemented and calculated and has high accuracy, which can work in the real time digital communication network with better signal-noise-ratio. The single frequency signal rapid detection, which has rapid response, would process the special audio signaling. These two algorithms are both used in the designed system with satisfactory effectiveness.Then, we put forward an optimized automatic gain control algorithm to fulfill the gain control based on many other speech enhancement methods. With the flatness of the gainvalue, we can get more impressive enhancement purpose.In the end, our automatic level control system is complemented under the criteria of ITU-T G. 169, and the Matlab simulation experiments demonstrate the effectiveness of our system.