针对机器人的应用场合通常存在各种噪声干扰的问题,提出了一种基于稀疏编码的语音特征提取方法.利用稀疏编码能稀疏表示语音的特性,在梅尔频域对语音增强后提取特征,将稀疏去噪与语音特征提取相融合,实现了混噪语音的有效补偿.在预设场景中的实验结果表明,与现有特征提取方法相比,所提出的语音特征提取方法能有效降低噪声对语音特征的影响,提高机器人语音控制的性能. Aiming at the problem of robots, there often exist various kinds of noise interference, and a speech feature extraction method based on sparse coding is proposed. Sparse coding can sparsely represent the characteristics of speech, extract features from mel frequency domain and extract The sparse denoising combines with the speech feature extraction to realize the effective compensation of the noisy speech.Experimental results in the preset scene show that compared with the existing feature extraction methods, the proposed speech feature extraction method can effectively reduce the noise pair Voice features, improve the performance of robot voice control.