基于SDCN算法的鲁棒性语音命令识别

提出了一种基于ＳＤＣＮ算法的鲁棒性语音命令识别。依赖于信噪比的倒谱正常化（ＳＤＣＮ）算法直接在倒谱域根据输入语音帧的信噪比（ＳＮＲ）来增加一补偿矢量，从而恢复未受污染的净语音信号，补偿矢量直接从训练环境和测试环境中记录的语音倒谱中逐帧比较得到，该算法对退化的环境具有很强的鲁棒性，实验结果证明，该算法简单，有效。

Robust Voice Command Recognition Based on SDCN Algorithm

A robust voice command recognition based on SDCN algorithm was presented. The SDCN (SNR Dependent Cepstral Normalization) algorithm operates directly in the cepstral domain by adding a compensation vector that depends on the SNR of the input frame, thus the uncorrupted voice can be recovered. These compensation vectors are computed from direct frame by frame comparisons of cepstrum of voice simultaneously recorded in the training environment and testing environments. This algorithm can perform good robustness under degraded environments, and the experimental results prove that the SDCN algorithm is simple and effective.