【摘要】 将稀疏分解引入到声目标识别系统的去噪中,提出了一种并行的稀疏分解算法,以解决目前的稀疏分解算法计算量大、耗时多,不适宜用作战场声目标实时识别的缺点。算法充分利用目前计算机高速发展的多核处理器技术,根据声信号的特点,将基于稀疏分解的声信号去噪分解成多核计算机可以并行处理的多个任务,使用加权的匹配跟踪算法分别进行处理,提高了运算速度。实验结果表明基于稀疏分解的去噪方法可以很好地提高声目标识别系统的性能,而并行的匹配跟踪算法可以显著提高稀疏分解的速度,使稀疏分解去噪应用到实时的声目标识别系统中成为可能。更多还原

【Abstract】 The sparse decomposition is pulled in to the de-noising in acoustic target recognition in this paper.Moreover,in order to overcome the deficiency of the computational burden and large time consuming,a parallel algorithm is put forward to take advantage of multi-core platforms of modern computer.We break down the decomposition into multiple local tasks according to the characteristic of acoustic signal while avoiding blocking effects.Experimental results show that the sparse decomposition de-noising can achieve significant advancement while the parallel MP algorithm reduces the time consuming,which make it possible to use sparse decomposition in real-time acoustic targets recognition system.更多还原