Researth And Application Of Speech Enhancement Algorithms

Noise exists everywhere in our modern life.The speech signal, in its collection or transmission processes, can’t avoid to be subject to the surrounding environment, the transmission media, the communication equipment, as well as the speaker himself. These disturbances will seriously affect the acceptance and the use of voice signals. Speech enhancement deals with the noisy speech to improve the quality of the speech, enhance speech’s clarity, and improve the performance of the speech signal applications. Therefore, speech enhancement algorithm has a very wide range of applications in our life.Aiming to deal with speech enhancement, this thesis tries to explore an effective algorithm, considering its flexibility, adaptability and practicality. Focusing on the research objectives, this thesispresents the main works as follows:Firstly, do some pre-treatment to the speech signal, including pre-emphasis, basic windowing, framing, and short analysis. And then, the difference between short-term features of the speech signal and the noise signal are fully analyzed. At last, finish the voice activity detection by the double threshold detection method, based on the difference between the two signals.Secondly, two parameters are introduced into the basic spectral subtraction. Using the enhanced signal’s SNR as the output, the two parameters as the input, this thesis choose the two optimal parameters automatically. What’s more, the selected optimal parameters are matched to the input SNR. When performing the enhancement processing, the two optimal parameters are got by calculating the input SNR according to the optimal parametric Curve. Then, smoothing process’ s combination will maximize the speech enhancement. The final simulation experiments demonstrate the superiority of the proposed algorithm.Finally, the improved speech enhancement algorithm is applied to the robot sound source localization system. This thesis select the tetrahedron sound source microphone array, and then establish the sound source localization model. By analyzing the positioning model,it shows the outstanding importance of the delay estimation accuracy.The final simulations and experiments verify the proposed algorithm effective. With its application, the sound source localization system has improve the TDE accuracy, and enhanced the positioning accuracy.