汽车驾驶性评价中小波去噪分解层数的确定

Determination of the Number of Wavelet Denoising Decomposition Layers in Vehicle Driving Evaluation

在汽车驾驶性评价试验过程中所采集的信号内通常混有诸多噪声信号,一般运用小波变换进行去噪,然而若小波去噪分解层数选择不当,则会影响去噪效果,从而影响驾驶性评价的准确性。针对此问题,提出了一种通过信息熵法融合所采集的汽车驾驶性数据的均方根误差、信噪比以及平滑度等多指标来选择小波分解最佳层数的方法,从而取得较佳的去噪效果。现对所采集的某一换挡工况下汽车振动加速度数据运用上述方法进行去噪,结果表明,基于此多指标融合的小波去噪方法在很好滤除初始信号中噪声的同时,能够有效保留数据中的有用成分,譬如驾驶过程中的冲击及振动等,从而保证了后续汽车驾驶性评价过程中所提取的指标的准确性。

In the process of vehicle driving evaluation test, a lot of noise signals are usually mixed in the collected signals. And the wavelet transform is usually used for denoising. However, if the number of the wavelet denoising decomposition layers is not properly selected, it will reduce the denoising effect and lower the accuracy of the driving evaluation. For this reason, a multi-index fusion method based on root mean square error, signal-to-noise ratio and data smoothness, collected and combined by information entropy method, is proposed to select the optimal number of layers for wavelet decomposition, so as to obtain the best denoising effect. This method is performed for the acceleration data denoising of the car’s vibration signals collected under a certain shifting condition. The results show that the wavelet denoising method based on this multi-index fusion can well filter the noise in the initial signal and retain the useful components such as various shocks and vibrations during driving so as to ensure the accuracy of the indicators extracted during the subsequent driving evaluation.