基于SOPC的混叠语音降噪方法

Study on Overlapping Speech Algorithm for Noise Reduction Based on SOPC

混叠语音干扰降低了语音识别系统、语音控制系统的性能,而且普通的频分技术无法将其去除。从提取语音特征入手,利用语音信号的非线性产生模型、能量分离算法,去除混叠在有用语音中的含绝大部分能量的元音干扰。最后尝试以SOPC技术结合NiosⅡ软核处理器,完成硬件实现的总体设计。

The overlapping speech interference reduces the performance of speech recognition system and speech control system,and the overlapping speech interference can not be removed by using an ordinary frequency division technology.The paper starts from extracting the speech feature,uses Non-linear generation model and ESA to remove the vocalic interference which contains the vast majority of energy in the useful speech.The paper also trys to use SOPC technology combined with NiosⅡ,to complete the overall hardware design.