Wearable Hearing Device Spectral Enhancement Driven by Non-Negative Sparse Coding-Based Residual Noise Reduction

Abstract

This paper proposes a novel technique to improve a spectral statistical filter for speech enhancement, to be applied in wearable hearing devices such as hearing aids. The proposed method is implemented considering a 32-channel uniform polyphase discrete Fourier transform filter bank, for which the overall algorithm processing delay is 8 ms in accordance with the hearing device requirements. The proposed speech enhancement technique, which exploits the concepts of both non-negative sparse coding (NNSC) and spectral statistical filtering, provides an online unified framework to overcome the problem of residual noise in spectral statistical filters under noisy environments. First, the spectral gain attenuator of the statistical Wiener filter is obtained using the a priori signal-to-noise ratio (SNR) estimated through a decision-directed approach. Next, the spectrum estimated using the Wiener spectral gain attenuator is decomposed by applying the NNSC technique to the target speech and residual noise components. These components are used to develop an NNSC-based Wiener spectral gain attenuator to achieve enhanced speech. The performance of the proposed NNSC–Wiener filter was evaluated through a perceptual evaluation of the speech quality scores under various noise conditions with SNRs ranging from -5 to 20 dB. The results indicated that the proposed NNSC–Wiener filter can outperform the conventional Wiener filter and NNSC-based speech enhancement methods at all SNRs.

Auditory filterbank denoising neural network for speech enhancement in wearable auditory device

“…As a fundamental loss function for calculating $${\mathcal{L}}_{P}$$ and $${\mathcal{L}}_{m}$$, the scale‐invariant signal‐to‐noise ratio (SI‐SNR) is utilized, which is known for its effective denoising performance [4, 5]. Here, $${\mathcal{L}}_{P}$$ is defined as follows: $$$$\begin{equation} \ {\mathcal{L}}_{p}=\ 10\textit{lo}{g}_{10}\left(\Vert {\mathbf{s}}^{\textit{targ}\textit{et}}{\Vert}_{2}^{2}/\Vert {\mathbf{e}}^{\textit{noise}}{\Vert}_{2}^{2}\right), \end{equation}$$$$where $${{{\bf s}}}^{target} = \frac{{\langle {{{\bf \hat{s}}}( {{{\bf n}} - \tau } ),{{\bf s}}( {{{\bf n}} - \tau } )} \rangle {{\bf s}}( {{{\bf n}} - \tau } )}}{{\|{{\bf s}}( {{{\bf n}} - \tau } )\|_2^2}}\ $$ and $${\mathbf{e}}^{\textit{noise}}=\widehat{\mathbf{s}}(\mathbf{n}-\tau )-{\mathbf{s}}^{\textit{targ}\textit{et}}$$.…”

“…Monaural auditory device algorithms, including filterbanks (FBs), should not introduce delays exceeding 10 ms, and computational complexity should be low to accommodate the limited computational capacity and battery power of portable devices [3]. To meet these criteria, the polyphase discrete Fourier transform (DFT) FB strategy is predominantly utilized, which is implemented through the integrated structure of fast Fourier transform (FFT)-based short-term Fourier transform (STFT) analysis and overlap-add techniques [3,4].…”

CNN-based noise reduction for multi-channel speech enhancement system with discrete wavelet transform (DWT) preprocessing

“…Microphone arrays and speech enhancement components are built into MCSE that processes multiple channels of audio signals in noisy environments such as outdoor environments ( Palla et al, 2017 ; Pauline, Samiappan & Kumar, 2021 ). For example, a spectral statistics filter is applied to hearing aids to handle stationary noise environments (Gaussian noise) and unsteady noise environments (factories, babble, and car noises) from −5 dB to 20 dB ( Kim, 2020 ). The current performance rate at low SNR are 2.16 PESQ score with babble noise, 2.20 considered as low quality of signal with Gaussian noise, 2.13 considered as low quality of signal with factory noise and 3.67 PESQ score considered as a medium quality of signal with car noise on an average of −5 dB to 10 db SNR levels ( Kim, 2020 ).…”

“…The existing MCSE provides speech recognition at a 71% Word Recognition Rate (WRR) at 10 dB SNR compared to a single microphone ( Xu et al, 2004 ; Stupakov et al, 2012 ). These multi-channel algorithms (beamforming, adaptive noise reduction and voice activity detection algorithms) suffer from the low performance of recognition rate when SNR is low (−15 dB, −10 dB, −5 dB, 0 dB) ( Pauline, Samiappan & Kumar, 2021 ; Kim, 2020 ). The existing algorithms developed for MCSE systems were only tested for white Gaussian stationary noise at 0 to 60 dB SNRs and were never tested for non-stationary environmental noises.…”

The Performance of Wearable Speech Enhancement System Under Noisy Environment: An Experimental Study

“…Microphone array and speech enhancement are the components embedded in wearable speech enhancement (WSE) that process speech signals in multichannel under noisy environments such as the outdoor environments [3]. For example, a spectral statistical filter is applied in wearable hearing devices for handling stationary noise environment (Gaussian noise) and non-stationary noise environments (babble, factory, and car) at −5dB to 20dB [4].…”