Sound quality estimation for nonstationary vehicle noises based on discrete wavelet transform

A new method based on the discrete wavelet transform (DWT) for sound quality estimation (SQE) of nonstationary vehicle noises is presented in this paper. Sample interior and exterior vehicle noises are first measured and denoised using the wavelet threshold method to initially lower the noise level with negligible signal distortion. A multirate filter bank combined by a set of low-pass and band-pass filters and a DWT-based filter bank are then developed for sound octave-band analysis (OBA). To build the DWT octave-band filters, the Mallat pyramidal algorithm is introduced, and five types of wavelet function with different filter lengths are investigated and compared. Finally, the Daubechies wavelet with a filter order of 35 is determined and applied to estimate sound pressure levels (SPLs) of both the interior and exterior vehicle noises. Verification results show that the newly proposed multirate-filter-based method (MF-OBA) and DWT-based method (DWT-OBA) are accurate and effective for SQE of nonstationary vehicle noises. Due to outstanding time-frequency characteristics of the wavelet analysis, the DWT-OBA can be suggested to deal with not only SQE of nonstationary vehicle noises, but also any other sound-related signal processing in engineering.