摘要
为了解决无线传感器网络采集到的车辆声信号易动态时变,易受外界环境影响,导致后续识别效果很差的问题,运用离散小波理论对其进行阈值去噪处理。采用"Daubechies4"小波,对信号进行5次分解,小波阈值函数采用软硬阈值折衷法,阈值系数的选取采用局部(分层)阈值,并利用XDS510仿真器在CCSLink中对算法进行编程,在主控芯片TMS320F2812器件中得以实验验证。实验肯定了运用离散小波算法能有效地去除混叠在车辆声信号中的噪声信号,有助于后续的识别工作,能够将对车辆声信号的识别率提高约30%。
The vehicle-acoustic signal sampled by wireless sensor network is liable to dynamic change and being influenced by environment, which results the bad identification effect. The discrete wavelet theory is adopted to estimate the signal noise, using ＂Daubechies4＂ wavelet to decompose the signal for 5 times, adopting soft and hard threshold compromise of wavelet thresholding function, using partial （stratified） threshold of threshold coefficient, and programming with XDSS10 emulator at CCSLink. The arithmetic is implemented in TMS320F2812. The test indicates this discrete wavelet arithmetic can mitigate the noise signal mixed in the vehicle-acoustic signal and improve the recognition rate of vehicle-acoustic signal by 30%.
出处
《信息与电子工程》
2009年第2期132-135,共4页
information and electronic engineering
基金
国家自然科学基金资助项目(60575027)
