Kwang Myung Jeon

A lossless compression method incorporating sensor fault detection for underwater acoustic sensor array

This article proposes an efficient lossless compression method for an underwater acoustic sensor array. The proposed method first decides whether the input signal is coming from a normal or a faulty sensor. In particular, such sensor fault detection is performed using a root-mean-square crossing rate and a zero crossing rate. If a sensor is determined to be faulty, then a pre-processing technique is applied prior to signal compression to make the signal from this sensor meaningful for further use. After that, multi-channel sensor signals are encoded by the MPEG-4 audio lossless coding encoder, where the indices of faulty sensors are also sent to the decoder, which applies a post-processing technique on the MPEG-4 audio lossless coding decoded signals for the faulty sensors. The performance of the proposed method is evaluated using three different sensor arrays deployed in real underwater environments by measuring the precision of faulty sensor detection and comparing the compression ratios between normal and faulty sensors. The evaluation shows that the fault detection of the proposed method works correctly for the experiments, and the compression ratios are decreased by 0.43% and 4.57%, respectively, in the faulty sensor signal send and non-send modes, compared to the MPEG-4 audio lossless coding reference software.

Audio enhancement using local SNR-based sparse binary mask estimation and spectral imputation

Abstract This paper proposes a method for enhancing speech and/or audio quality under noisy conditions. The proposed method first estimates the local signal-to-noise ratio (SNR) of the noisy input signal via sparse non-negative matrix factorization (SNMF). Next, a sparse binary mask (SBM) is proposed that separates the audio signal from the noise by measuring the sparsity of the pool of local SNRs from the adjacent frequency bands of the current and several previous frames. However, some spectral gaps remain across frequency bands after applying the binary masks, which distorts the separated audio signal due to spectral discontinuity. Thus, a spectral imputation technique is used to fill the empty spectrum of the frequency band where it is removed by the SBM. Spectral imputation is conducted by online learning NMF with the spectra of the neighboring non-overlapped frequency bands and their local sparsity. The effectiveness of the proposed enhancement method is demonstrated on two different tasks use speech and musical content, respectively. Consequently, objective measurements and subjective listening tests show that the proposed method outperforms conventional speech and audio enhancement methods, such as SNMF-based alternatives and deep recurrent neural networks for speech enhancement, block thresholding, and a commercially available software tool for audio enhancement.