This paper proposes a speech enhancement method using the multi-scales and multi-thresholds of the auditory perception wavelet transform, which is suitable for a low SNR (signal to noise ratio) environment. This method achieves the goal of noise reduction according to the threshold processing of the human ear's auditory masking effect on the auditory perception wavelet transform parameters of a speech signal. At the same time, in order to prevent high frequency loss during the process of noise suppression, we first make a voicing decision based on the speech signals. Afterwards, we process the unvoiced sound segment and the voiced sound segment according to the different thresholds and different judgments. Lastly, we perform objective and subjective tests on the enhanced speech. The results show that, compared to other spectral subtractions, our method keeps the components of unvoiced sound intact, while it suppresses the residual noise and the background noise. Thus, the enhanced speech has better clarity and intelligibility.
In this paper, a new lifting wavelet domain audio watermarking algorithm based on the statistical characteristics of sub-band coefficients is proposed. First of all, an original audio signal was segmented and each segment was divided into two sections. Then, the Barker code was used for synchronization, the LWT (lifting wavelet transform) was performed on each section, a synchronization code and a watermark were embedded into the first section and the second section, respectively, by modifying the statistical average value of the sub-band coefficients. The embed strength was determined adaptively according to the auditory masking property. Experiments show that the embedded watermark has better robustness against common signal processing attacks than present algorithms based on LWT and can resist random cropping in particular.
The normal modes cannot be extracted even in the Pekeris waveguide when the source-receiver distance is very close. This paper introduces a normal mode extraction method based on a dedispersion transform (DDT) to solve this problem. The method presented here takes advantage of DDT, which is based on the waveguide invariant such that the dispersion associated with all of the normal modes is removed at the same time. After performing DDT on a signal received in the Pekeris waveguide, the waveform of resulting normal modes is very close to the source signal, each with different position and amplitude. Each normal mode can be extracted by determining its position and amplitude parameters by applying particle swarm optimization (PSO). The waveform of the extracted normal mode is simply the waveform of the source signal; the real waveform of the received normal mode can then be recovered by applying dispersion compensation to the source signal. The method presented needs only one receiver and is verified with experimental data