引用本文: |
-
李嘉新,魏爽,俞守庚,等.基于字典尺度自适应学习的欠定盲语音重构算法[J].电讯技术,2023,63(9): - . [点击复制]
- LI Jiaxin,WEI Shuang,YU Shougeng,et al.An Underdetermined Blind Speech Reconstruction Algorithm Based on Adaptive Scale Dictionary Learning[J].,2023,63(9): - . [点击复制]
|
|
摘要: |
针对欠定盲语音分离传统字典学习算法不能优化字典尺寸的问题,提出了一种尺度自适应同步码字优化(Scale Adaptive Simultaneous Codeword Optimization,SASimCO)算法。设计了一种迭代调整字典尺寸的自适应字典学习策略,将训练的字典用于语音盲分离中,以提高语音源信号的恢复性能。所提算法依据设计的候选矩阵,计算候选矩阵中的原子重要性,按照原子重要性准则对字典进行添加与删除原子操作,最后迭代训练得到一个稀疏表示误差最优的字典,用于语音源信号的恢复。使用SiSEC(Signal Separation Evaluation Campaign)数据集对所提算法进行的仿真实验表明,相较于传统字典学习算法,所提算法提高了1~3 dB语音源分离性能,证明了该算法的优势。 |
关键词: 欠定盲源分离 语音重构 尺度自适应字典学习 稀疏表示 |
DOI:10.20079/j.issn.1001-893x.220302004 |
|
基金项目: |
|
An Underdetermined Blind Speech Reconstruction Algorithm Based on Adaptive Scale Dictionary Learning |
LI Jiaxin,WEI Shuang,YU Shougeng,LIU Rui |
(1.College of Information,Mechanical and Electrical Engineering,Shanghai Normal University,Shanghai 201418,China;2.Institute of Sensing and Navigation,Shanghai Jiaotong University,Shanghai 200030,China) |
Abstract: |
In underdetermined blind speech separation,a Scale Adaptive Simultaneous Codeword Optimization(SASimCO) algorithm is proposed for the problem that traditional dictionary learning algorithms cannot optimize the dictionary size.An adaptive dictionary learning strategy that iteratively adjusts the dictionary size is designed and these dictionaries with optimal size in speech blind separation are used to improve the recovery performance of the speech source signals.According to the designed candidate matrix,the importance of the atoms in the candidate matrix is calculated.According to the atom importance criterion,the atoms are added to or removed from the dictionary.After iteratively training,an optimal dictionary with the least sparse representation error is obtained for the recovery of speech source signals.Simulation with the Signal Separation Evaluation Campaign(SiSEC) dataset shows that,compared with the traditional dictionary learning algorithms,the proposed algorithm improves the speech source separation performance by 1~3 dB,which proves the advantage of the algorithm. |
Key words: underdetermined blind source separation speed reconstruction scale adaptive dictionary learning sparse representation |