The Research On Technique Of Train Speech Signal Noise Reduction And Optimization Of G.726 Algorithm

The train monitoring and controlling device was proposed upon experiences of trains' operation, the voice recording and compressing part included. The main functions of the voice recording and compressing part are to record speech signal of train drivers. The record could be played by administrations, such as mechanical engineering department and electric engineering department. Reappearance of sound of train and speech signal of train drivers' is important for those administrations since they can grasp conditions of the train and drivers of that time. Then they would detect defects of hardware and managements. In case of train accidents, it is important to analyze the reason and experience. By this way, they will move amendments for devices and managements of trains, and responsibility of accidents included. Therefore, study of speech sound recording device appeared in this paper.In this paper, the first task is to introduce history and development of voice handling technology. Secondly, the train voice-handling project was introduced and speaker recognition algorithm was chosen based on conditions of trains. Thirdly, the keynote cycle estimation under train noise condition is given in this paper, the keynote cycle estimation had take train noise into account. Last, noise reduction algorithms and theirs emulation added in this paper, they are put forward based on real condition of train, such as train noise and real-time-quality. In the simulation of noise reduction algorithms, we found they did it well. In voice-spectrum picture we can see it clearer. Next, algorithm G.726 and its optimization were explained .In the simulation of the optimized algorithm, little quantization-noise appeared. It does not matter since our ears are not so sensitive that we feel it noisy. In this paper, cut down operand and real time capability were taken into account.