| In our daily life, the speech signal is inevitably affected by the noises in the process of production, transmission and reception. Noise can degrade the performance of speech processing system and affect the normal speech communication. In order to improve the quality of the noisy signal, we need to reduce the noise. Speech enhancement is the effective method to solve noise pollution, the main target of which is to extract the original speech signal as pure as possible from the corrupted speech signal in the receiver and thus improve the speech definition and intelligibility.Speech enhancement under strong noise background is mainly discussed in the speech enhancement system of this thesis. According to the research content, there are two logical parts of this dissertation.The first part is algorithm research and simulation. Firstly, the basic theory of speech signal processing which is the foundation of the speech enhancement algorithm is presented. Then, we give a detailed discussion of the speech enhancement algorithm based on short-time spectrum estimation, including spectrum subtraction and STSA-MMSE algorithm. Experiment results show the improved performance of the two methods. Additionally, this article also introduced independent component analysis (ICA) method. The fast independent component analysis (FICA) algorithm is analyzed concretely. Finally, all algorithms are applied to the simulation experiments to analyze and compare to their performances.The second part concentrates on the construction of DSP platform and algorithm migration. Through above simulation tests, we decide to apply the FICA algorithm to DSP In the speech enhancement system of this essay, TMS320VC5509 chip of the TI company is used to build hardware platform. This system includes the data acquisition module, data processing and data transmission module. At last, the experimental results indicate that the system has excellent real time processing capacity and satisfactory speech enhancement results. |
PDF Full Text Request