Low-Power End-to-End Cochlear Implant Speech Denoising with Spiking Neural Networks

Ludovic Boulanger; Sean U. N. Wood

cs.SDAug 28, 2026

Low-Power End-to-End Cochlear Implant Speech Denoising with Spiking Neural Networks

Authors: Ludovic Boulanger, Sean U. N. Wood

Organizations: Department of Electrical and Computer Engineering University of Sherbrooke, QC, Canada

Abstract

Cochlear implants (CI) restore hearing for individuals with severe to profound hearing loss. However, CI users often struggle to understand speech in noisy environments. Deep neural networks (DNN) have shown promise in enhancing speech for CI users, yet their high energy demands make them non-ideal for low-power CI processors. Spiking neural networks (SNN), on the other hand, offer comparable performance with significantly lower energy consumption. Hence, we propose a novel SNN inspired by the Deep ACE architecture that simultaneously performs speech enhancement and CI coding. Our model achieves competitive vocoded short-time objective intelligibility (VSTOI) and signal-to-noise ratio improvement (SNRi) scores compared to Deep ACE, while achieving more than a sixfold reduction in energy consumption.

Figures & tables

Figure 1: A) Spiking Deep ACE processing pipeline. B) Deep ACE processing pipeline. C) Convolutional blocks used in Spiking Deep ACE. D) Convolutional blocks used in Deep ACE.

Figure 2: Example of predicted electrodograms by each of the models as well as the clean and noisy electrodograms for reference.

Hyperparameter	Deep ACE	Spiking Deep ACE
N	64	128
L	32	32
B	64	64
Sc	32	64
H	128	64
P	3	3

Table 1: Hyperparameters used in Deep ACE and Spiking Deep ACE.

Panel of Figure 3 — Figure 3: a) Box plots showing the VSTOI scores for both models on the ICRA babble and ICRA static test sets. b) Box plots showing the SNRi scores for both models on the same test sets as in a). In both a) and b), the center bar in each box represents the median score. The top and bottom edges of the box represent the third and first quartile respectively. The box whiskers extend to 1.5 times the inter-quartile range (IQR). Circles indicate results outside of 1.5 times the IQR.

Figure 4: Evolution of the mean spike rate per neuron during training. Since the sampling rate of the network is $1\text{\,}\mathrm{kHz}$ , a spike rate of 250 spikes / second implies one spike every 4 time steps.

Model	VSTOI (%)	SNRi (dB)	Energy ( $\mu J/s$ )
Spiking Deep ACE	56	6.0	372
Deep ACE	57	6.1	2461

Table 2: Results and energy comparison between Deep ACE and Spiking Deep ACE on the ICRA babble and ICRA static test sets combined.

Speech Processing

Speech processing develops computational methods for recognizing, synthesizing, analyzing, and transforming spoken audio.

See all 410 papers

DBHN-Net: Dual-Branch Hybrid Neural Network For Low-Complexity Monaural Speech Enhancement
Jun 4, 2026Cunhang Fan, Enrui Liu, Jing Zhou +6Efficient Neural Network Inference Speech Enhancement
State Key Laboratory of Opto-Electronic Information Acquisition and Protection Technology, (School of Computer Science and Technology), Anhui University, Hefei, 230601, Anhui, P. R. China · China Telecom Artificial Intelligence Technology (Beijing) Co., Ltd · Institute of Acoustics, University of Chinese Academy of Sciences, Beijing 100190, China +1
Neuromorphic Speech Enhancement with Dual-Branch Spiking Neural Networks
Jun 22, 2026Taiyu Meng, Wenbin Jiang, Haoyi Zhang +2Speech Enhancement Spiking Neural Networks
School of Communication Engineering, Hangzhou Dianzi University, Hangzhou, China
Adaptive Speech-to-Spike Encoding for Spiking Neural Networks
Jun 17, 2026Taharim Rahman Anon, Jakaria Islam EmonKeyword Spotting Speech Processing
1PI LLC, Sapporo, Hokkaido, Japan

Jun 4, 2026J/K move · Enter open · S save

DBHN-Net: Dual-Branch Hybrid Neural Network For Low-Complexity Monaural Speech Enhancement

Cunhang Fan, Enrui Liu, Jing Zhou, Jian Kang+5

State Key Laboratory of Opto-Electronic Information Acquisition and Protection Technology, (School of Computer Science and Technology), Anhui University, Hefei, 230601, Anhui, P. R. China · China Telecom Artificial Intelligence Technology (Beijing) Co., Ltd · Institute of Acoustics, University of Chinese Academy of Sciences, Beijing 100190, China+1

Efficient Neural Network Inference Speech Enhancement Spiking Neural Networks

Open paper AskPDF arXiv

Although artificial neural network (ANN) based speech enhancement (SE) methods demonstrate excellent performance, the high computational complexity and high energy consumption hinder their deployment in practical front-end processing tasks.} Currently, the spiking neural networks (SNNs) have shown potential in reducing power consumption. However, the discrete binary activation and complex spatio-temporal dynamics of SNNs often result in information loss. The current challenge therefore focuses on how to maintain performance and reduce computational complexity. To address this issue, this work propose a Dual-Branch Hybrid Neural (DBHN) Network. 1) In terms of network architecture: A dual-branch network integrating ANN and SNN was designed, where the SNN branch reduces power consumption while the ANN branch addresses information loss; The BandSplit and Time-Frequency (TF) -Mamba modules were developed to simultaneously compress energy consumption and enhance model performance; Spiking Feature Extraction Group (SFEG) and Information Transformation Block (ITB) components were implemented with residual connections to mitigate information loss while further refining feature representations. 2) To facilitate inter-branch information fusion: An Interaction module was designed to promote information exchange at various stages of the dual-branch network; A TF-Cross Attention-Fusion module was designed to perform time-frequency domain fusion of dual-branch information while data-adaptively guiding the SNN branch to retain more critical information. Results show that the proposed model maintains superior performance across three public datasets while achieving an average 7.5 fold reduction in computational complexity compared to baseline models.

Abstract

Speech Processing

Explore similar work

DBHN-Net: Dual-Branch Hybrid Neural Network For Low-Complexity Monaural Speech Enhancement

Neuromorphic Speech Enhancement with Dual-Branch Spiking Neural Networks

Adaptive Speech-to-Spike Encoding for Spiking Neural Networks