Study Of Near-field Microphone Arrays Based Speech Enhancement Algorithm

Microphone array is widely used in a variety of electronic products. Microphone array-based systems have better performance than that base on single microphone. In fact, the mostused microphone array speech enhancement algorithm, which is base on the far-field model,is not optimal. People are usually closed to the Microphone array, when they use it. They areso closed that the far-field model assumptions are not satisfied. Therefore, we need anavailable near-field model based algorithm to replace the far-field model based on. Then wewould get a more effectively microphone array system. This paper is mainly about thealgorithm of the near-field based microphone array speech enhancement.Firstly, I introduced the principle of the microphone array speech enhancementalgorithms. Then I introduced the the near-field model and far-field model, and the similaritiesand differences between them. Make the necessary preparations for the following discussion.Secondly, I introduced far-field based and near-field based DOA algorithm. Clarify theprinciples and importance of the direction of arrival estimation algorithm. Discussed thenear-field based MUSIC algorithm and MUSIC-like algorithm. Then made simulations ofthem.Thirdly, I introduce of how to use MSNR, MSE and LCMV base on the near-field model.Made simulations of them, and drawn the conclusion.According to the MSNR and LCMV, I propose a new beamformer which get the bothadvantage of MSNR and LCMV. The new a algorithm is low time complexity. It run betterthan MSNR and LCMV in actual system.Finally, I drawn a concluding. And something about the further of the near-field modelbased algorithm research.