Abstract:

The conventional audio transmission based on 1-bit Compressed Sensing（CS） suffers the problems of poor reconstruction accuracy and sound quality. In this paper, a 1-bit reconstruction method incorporating the featured information of sparse audio signal is proposed. At the sending-end, the featured information, which is constructed with the partial support set of the sparse audio signal, is superimposed on the 1-bit compressed audio signal for further transmission. At the receiving-end, the feature information and the 1-bit compressed audio signal are recovered, and the feature-aided 1-bit reconstruction algorithm is developed for audio reconstruction. Compared with the conventional audio reconstruction method based on 1-bit CS, the proposed method can improve the Mean Square Error（MSE） and Mean Opinion Score（MOS） of the reconstructed audio without increasing the cost of spectrum resources.

Key words: audio transmission, 1-bit compressed sensing, feature aided, superimposed sequence

摘要：

针对1-bit压缩感知音频传输存在重构精度与音频质量较差的问题，提出稀疏音频信号特征信息辅助的1-bit重构的方法。发送端利用稀疏音频信号的部分支撑集构建特征信息，并将特征信息扩频后叠加到1-bit压缩的音频信号上传输；接收端恢复特征信息和1-bit压缩的音频信号，并构建特征辅助的重构算法以恢复音频信号。相较于经典的1-bit音频压缩重构方法，所提方法可在不增加频谱开销的情况下改善恢复音频的MSE（Mean Square Error）值和MOS（Mean Opinion Score）评分。

关键词: 音频传输, 1-bit压缩感知, 特征辅助, 叠加序列