非平稳信号稀疏表示的研究发展

信号分解是从信号中获取特征信息的过程，是模式识别、智能系统和故障诊断等诸多领域的基础和关键。非平稳信号往往包含着反映系统变化的重要信息，并且广泛存在，对其研究具有非常重要的理论意义和工程应用价值。以改进信号表示的稀疏性为主线，分析了推动非平稳信号特征提取方法发展的工程背景，详细描述了5类特征提取方法的特性与机理、历史沿革和面临的挑战，比较研究了各种方法的模型，并系统评述了这些模型在信号处理和分析中的最新进展，以及在一些领域中的应用。最后指出了各种方法目前存在的问题和不足，探讨了进一步的研究重点。

Recent advances in sparse representation of non-stationary signal

Signal decomposition is a process that obtains information from signals and it is a foundational and key technique for many fields such as pattern recognition, intelligent system and machinery fault diagnosis. It is very important to study non-stationary signal decomposition which always includes lots of information that can reflect the changing of the system and widely exists. After improving the sparsity of signal representation, the engineering background of feature extraction for non-stationary signal was studied in this paper, the characteristics, mechanisms, development history and current and future challenges of five types of methods were analyzed in depth, the models of these methods were compared, together with the state-of-the-art of feature extraction models in signal processing and analysis and some successful applications available were systematically reviewed. Finally, several main problems and a few deficiencies were pointed out, and future research directions were anticipated.