Analysis and implementation of low‐power perceptual multiband noise reduction for the hearing aids application

Abstract

Traditional noise reduction designs provide good performance but suffer from high complexity and long latency, which limits their application to hearing aids. Targeted for strict low-power and low-latency requirement of completely-in-the-canal type hearing aids, this study analyses and implements a previously proposed sample-based perceptual multiband spectral subtraction with a multiplication-based entropy voice activity detection. Simulation results reveal that the authors design can provide similar speech quality as others, but with lower computational complexity and simple control effort. The corresponding core-based architecture design further exploits processing characteristics of the proposed approach to reduce power consumption with a sign-magnitude and a preprocessed input data reuse scheme. Chip measurement shows that the design only consumes 83.7 µW at 0.6 V operation with 90 nm high threshold voltage (HVT) (high V T) standard cell library.

An Improved Perceptual MBSS Noise Reduction with an SNR-Based VAD for a Fully Operational Digital Hearing Aid

“…The proposed NRA is implemented using frame-based FFT processing with a frame size of 320 samples (zero-padded to 512 samples for the FFT) corresponding to 20 ms, and a frame advance of 160 samples (corresponding to 10 ms). Figure 4 displays the accuracy of the proposed SNR-based VAD compared with the sample-based entropy VAD [28], [29] and FFT-based entropy VAD [40] at five various input SNR levels and under four types of background noise environments. The sample-based VAD [28], [29] proposes a feature parameter, called spectral entropy, which exploits the use of the banded structure on speech spectrogram.…”

“…Figure 4 displays the accuracy of the proposed SNR-based VAD compared with the sample-based entropy VAD [28], [29] and FFT-based entropy VAD [40] at five various input SNR levels and under four types of background noise environments. The sample-based VAD [28], [29] proposes a feature parameter, called spectral entropy, which exploits the use of the banded structure on speech spectrogram. And it adaptively selects useful bands not corrupted by noise to distinguish a speech from a nonspeech.…”

“…And the complicated logarithms and divisions are replaced by the multiplications according to the simplification method introduced in [29]. Moreover, the exponents are implemented in table lookup.…”

“…To evaluate the reduction performance (evaluated by segSNR) and speech quality (evaluated by PESQ) respectively, the proposed NRA is compared with other different noise reduction methods such as Wiener filter (wiener wt) [11], subspace (klt) [13], the mband [18], the sample-based [28], [29], and MBSC [30] at five different input SNR levels (0, 3, 6, 9, 12 dB) and under four types of background noise environments (white, babble, factory and car). The programs of the wiener wt, klt, and mband algorithms are provided by [2], and the programs of the samplebased and MBSC methods are provided by [28], [29] and [30], respectively.…”

“…The noise reduction method based on MBSS and entropy-based VAD was introduced Copyright c 2017 The Institute of Electronics, Information and Communication Engineers in 2014. This method can be simplified and still achieves good performance [28], [29]. In 2016, a speech enhancement approach based on analysis-synthesis framework was proposed [30].…”