摘要
传统的端点检测在低信噪比(SNR)非平稳噪声下性能会失效,因此文章提出了将最优改进的对数谱幅度估计(OMLSA)以及最小控制递归平均算法(IMCRA)相结合的方法对包含噪声的语音指令进行去噪处理,提取PNCC的第一维静态特征作为特征参数。同时,文章在单参数双门限法的基础上设计了一个自适应阈值,可以更好地跟踪预测实际语音的起始与终止端。Matlab仿真结果显示,该算法在各种非平稳噪声下比经典算法优势更大。
The performance of traditional endpoint detection will fail under low SNR non-stationary noise.Therefore,this paper proposes a method combining the optimal improved logarithmic Spectral Amplitude Estimation(OMLSA)and the Minimum Control Recursive Averaging algorithm(IMCRA)to denoise speech commands containing noise.The first dimension static features of PNCC are extracted as the feature parameters.At the same time,an adaptive threshold is designed based on the single-parameter double-threshold method,which can better track and predict the start and end of the actual speech.Matlab simulation results show that the proposed algorithm has more advantages than the classical algorithm under various non-stationary noises.
作者
高磊
章小兵
Gao Lei;Zhang Xiaobing(School of Electrical and Information Engineering,Anhui University of Technology,Ma’anshan 243032,China)
出处
《无线互联科技》
2023年第6期111-114,共4页
Wireless Internet Technology
基金
安徽工业大学产学研基金资助重大项目,项目编号:RD14206003。
