| 武朋辉,杨百龙,时磊.基于离散余弦变换的语音压缩采样和编码算法[J].,2015,34(1):17-23 |
| 基于离散余弦变换的语音压缩采样和编码算法 |
| Speech compressive sensing and codec algorithm based on DCT |
| 投稿时间：2014-03-28 修订日期：2014-12-25 |
| 中文摘要: |
| 针对语音无线通信中带宽资源受限的问题，提出基于压缩采样的低速率语音编码算法。以基尼系数为指标，比较不同稀疏变换域下语音信号的稀疏性，分析常见重构算法对语音信号压缩采样观测信号的重构特性。对标准耳蜗滤波器——伽马啁啾滤波器组的参数进行研究，并以梯度投影稀疏重建（GPSR）算法重构语音信号。利用语音质量感知评估（PESQ）、信噪比和主观听觉测试，对编解码后的合成语音信号进行了质量评估。实验表明，基于压缩感知的语音编码器以4kbps的低速率对语音进行编码时，PESQ得分可达到3.16，计算复杂度相对较低，可以用于实际的语音编码环境。 |
| 英文摘要: |
| Due to restricted bandwidth in wireless speech communication, a new low-bit rate speech codec based on compressive sampling under discrete cosine transform is proposed. Speech sparsity under different transformations is compared. The Gini index is utilized to gage the coefficient sparsity. Before sampling, gammachirp filterbank parameters are selected in the speech frame. During reconstruction, the GPSR is used to recover the signal. Subjective and objective tests show that the proposed technique gets 3.16 PESQ mean score, and the bit-rate reaches to 4kbps. Furthermore, compared to MELP, its computation complexity is relatively low and it can be used under real applications. |
| DOI：10.11684/j.issn.1000-310X.2015.01.003 |
| 中文关键词: 低速率编码，压缩采样，基尼系数，离散余弦变换 |
| 英文关键词: Low bit-rate coding,Compressive sampling,Gini index,DCT |
| 基金项目:军队装备科研基金资助项目 |
|
| 摘要点击次数: 2678 |
| 全文下载次数: 3703 |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
