This paper presents SHTNet, a lightweight spherical harmonic transform (SHT) based framework, which is designed to address cross-array generalization challenges in multi-channel automatic speech recognition (ASR) through three key innovations. First, SHT based spatial sound field decomposition converts microphone signals into geometry-invariant spherical harmonic coefficients, isolating signal processing from array geometry. Second, the Spatio-Spectral Attention Fusion Network (SSAFN) combines coordinate-aware spatial modeling, refined self-attention channel combinator, and spectral noise suppression without conventional beamforming. Third, Rand-SHT training enhances robustness through random channel selection and array geometry reconstruction. The system achieves 39.26\% average CER across heterogeneous arrays (e.g., circular, square, and binaural) on datasets including Aishell-4, Alimeeting, and XMOS, with 97.1\% fewer computations than conventional neural beamformers.
Although deep learning has substantially advanced speech separation in recent years, most existing studies continue to prioritize separation quality while overlooking computational efficiency, an essential factor for low-latency speech processing in real-time applications. In this paper, we propose SepPrune, the first structured pruning framework specifically designed to compress deep speech separation models and reduce their computational cost. SepPrune begins by analyzing the computational structure of a given model to identify layers with the highest computational burden. It then introduces a differentiable masking strategy to enable gradient-driven channel selection. Based on the learned masks, SepPrune prunes redundant channels and fine-tunes the remaining parameters to recover performance. Extensive experiments demonstrate that this learnable pruning paradigm yields substantial advantages for channel pruning in speech separation models, outperforming existing methods. Notably, a model pruned with SepPrune can recover 85% of the performance of a pre-trained model (trained over hundreds of epochs) with only one epoch of fine-tuning, and achieves convergence 36$\times$ faster than training from scratch. Code is available at https://github.com/itsnotacie/SepPrune.
In this paper, we propose a novel attention module termed the Differentiable Channel Selection Attention module, or the DCS-Attention module. In contrast with conventional self-attention, the DCS-Attention module features selection of informative channels in the computation of the attention weights. The selection of the feature channels is performed in a differentiable manner, enabling seamless integration with DNN training. Our DCS-Attention is compatible with either fixed neural network backbones or learnable backbones with Differentiable Neural Architecture Search (DNAS), leading to DCS with Fixed Backbone (DCS-FB) and DCS-DNAS, respectively. Importantly, our DCS-Attention is motivated by the principle of Information Bottleneck (IB), and a novel variational upper bound for the IB loss, which can be optimized by SGD, is derived and incorporated into the training loss of the networks with the DCS-Attention modules. In this manner, a neural network with DCS-Attention modules is capable of selecting the most informative channels for feature extraction so that it enjoys state-of-the-art performance for the Re-ID task. Extensive experiments on multiple person Re-ID benchmarks using both DCS-FB and DCS-DNAS show that DCS-Attention significantly enhances the prediction accuracy of DNNs for person Re-ID, which demonstrates the effectiveness of DCS-Attention in learning discriminative features critical to identifying person identities. The code of our work is available at https://github.com/Statistical-Deep-Learning/DCS-Attention.
Spike sorting is a valuable tool in understanding brain regions. It assigns detected spike waveforms to their origins, helping to research the mechanism of the human brain and the development of implantable brain-machine interfaces (iBMIs). The presence of noise and artefacts will adversely affect the efficacy of spike sorting. This paper proposes a framework for low-cost and real-time implementation of deep spike detection, which consists of two one-dimensional (1-D) convolutional neural network (CNN) model for channel selection and artefact removal. The framework utilizes simulation and hardware layers, and it applies several low-power techniques to optimise the implementation cost of a 1-D CNN model. A compact CNN model with 210 bytes memory size is achieved using structured pruning, network projection and quantization in the simulation layer. The hardware layer also accommodates various techniques including a customized multiply-accumulate (MAC) engine, novel fused layers in the convolution pipeline and proposing flexible resource allocation for a power-efficient and low-delay design. The optimized 1-D CNN significantly decreases both computational complexity and model size, with only a minimal reduction in accuracy. Classification of 1-D CNN on the Cyclone V 5CSEMA5F31C6 FPGA evaluation platform is accomplished in just 16.8 microseconds at a frequency of 2.5 MHz. The FPGA prototype achieves an accuracy rate of 97.14% on a standard dataset and operates with a power consumption of 2.67mW from a supply voltage of 1.1 volts. An accuracy of 95.05% is achieved with a power of 5.6mW when deep spike detection is implemented using two optimized 1-D CNNs on an FPGA board.
Vision foundation models (VFMs) are large pre-trained models that form the backbone of various vision tasks. Fine-tuning VFMs can further unlock their potential for downstream tasks or scenarios. However, VFMs often contain significant feature redundancy, which may limit their adaptability to new tasks. In this paper, we investigate the redundancies in the segment anything model (SAM) and then propose a parameter-free fine-tuning method to address this issue. Unlike traditional fine-tuning methods that adjust parameters, our method emphasizes selecting, reusing, and enhancing pre-trained features, offering a new perspective on model fine-tuning. Specifically, we introduce a channel selection algorithm based on the model's output difference to identify redundant and effective channels. By selectively replacing the redundant channels with more effective ones, we filter out less useful features and reuse the more relevant features to downstream tasks, thereby enhancing the task-specific feature representation. Experiments on both out-of-domain and in-domain datasets demonstrate the efficiency and effectiveness of our method. Notably, our approach can seamlessly integrate with existing fine-tuning strategies (e.g., LoRA, Adapter), further boosting the performance of already fine-tuned models. Moreover, since our channel selection involves only model inference, our method significantly reduces computational and GPU memory overhead.
Training on edge devices enables personalized model fine-tuning to enhance real-world performance and maintain data privacy. However, the gradient computation for backpropagation in the training requires significant memory buffers to store intermediate features and compute losses. This is unacceptable for memory-constrained edge devices such as microcontrollers. To tackle this issue, we propose a training acceleration method using dynamic gradient sparse updates. This method updates the important channels and layers only and skips gradient computation for the less important channels and layers to reduce memory usage for each update iteration. In addition, the channel selection is dynamic for different iterations to traverse most of the parameters in the update layers along the time dimension for better performance. The experimental result shows that the proposed method enables an ImageNet pre-trained MobileNetV2 trained on CIFAR-10 to achieve an accuracy of 85.77\% while updating only 2\% of convolution weights within 256KB on-chip memory. This results in a remarkable 98\% reduction in feature memory usage compared to dense model training.
We present OuroMamba, the first data-free post-training quantization (DFQ) method for vision Mamba-based models (VMMs). We identify two key challenges in enabling DFQ for VMMs, (1) VMM's recurrent state transitions restricts capturing of long-range interactions and leads to semantically weak synthetic data, (2) VMM activations exhibit dynamic outlier variations across time-steps, rendering existing static PTQ techniques ineffective. To address these challenges, OuroMamba presents a two-stage framework: (1) OuroMamba-Gen to generate semantically rich and meaningful synthetic data. It applies contrastive learning on patch level VMM features generated through neighborhood interactions in the latent state space, (2) OuroMamba-Quant to employ mixed-precision quantization with lightweight dynamic outlier detection during inference. In specific, we present a thresholding based outlier channel selection strategy for activations that gets updated every time-step. Extensive experiments across vision and generative tasks show that our data-free OuroMamba surpasses existing data-driven PTQ techniques, achieving state-of-the-art performance across diverse quantization settings. Additionally, we implement efficient GPU kernels to achieve practical latency speedup of up to 2.36x. Code will be released soon.
Learning-based algorithms have gained great popularity in communications since they often outperform even carefully engineered solutions by learning from training samples. In this paper, we show that the selection of appropriate training examples can be important for the performance of such learning-based algorithms. In particular, we consider non-linear 1-bit precoding for massive multi-user MIMO systems using the C2PO algorithm. While previous works have already shown the advantages of learning critical coefficients of this algorithm, we demonstrate that straightforward selection of training samples that follow the channel model distribution does not necessarily lead to the best result. Instead, we provide a strategy to generate training data based on the specific properties of the algorithm, which significantly improves its error floor performance.
Vehicle-to-everything (V2X) communication technology is revolutionizing transportation by enabling interactions between vehicles, devices, and infrastructures. This connectivity enhances road safety, transportation efficiency, and driver assistance systems. V2X benefits from Machine Learning, enabling real-time data analysis, better decision-making, and improved traffic predictions, making transportation safer and more efficient. In this paper, we study the problem of joint channel selection, where vehicles with different technologies choose one or more Access Points (APs) to transmit messages in a network. In this problem, vehicles must learn a strategy for channel selection, based on observations that incorporate vehicles' information (position and speed), network and communication data (Signal-to-Interference-plus-Noise Ratio from past communications), and environmental data (road type). We propose an approach based on Federated Deep Reinforcement Learning (FedDRL), which enables each vehicle to benefit from other vehicles' experiences. Specifically, we apply the federated Proximal Policy Optimization (FedPPO) algorithm to this task. We show that this method improves communication reliability while minimizing transmission costs and channel switches. The efficiency of the proposed solution is assessed via realistic simulations, highlighting the potential of FedDRL to advance V2X technology.
Epileptic seizures cause abnormal brain activity, and their unpredictability can lead to accidents, underscoring the need for long-term seizure prediction. Although seizures can be predicted by analyzing electroencephalogram (EEG) signals, existing methods often require too many electrode channels or larger models, limiting mobile usability. This paper introduces a SlimSeiz framework that utilizes adaptive channel selection with a lightweight neural network model. SlimSeiz operates in two states: the first stage selects the optimal channel set for seizure prediction using machine learning algorithms, and the second stage employs a lightweight neural network based on convolution and Mamba for prediction. On the Children's Hospital Boston-MIT (CHB-MIT) EEG dataset, SlimSeiz can reduce channels from 22 to 8 while achieving a satisfactory result of 94.8% accuracy, 95.5% sensitivity, and 94.0% specificity with only 21.2K model parameters, matching or outperforming larger models' performance. We also validate SlimSeiz on a new EEG dataset, SRH-LEI, collected from Shanghai Renji Hospital, demonstrating its effectiveness across different patients. The code and SRH-LEI dataset are available at https://github.com/guoruilu/SlimSeiz.
The taste electroencephalogram (EEG) evoked by the taste stimulation can reflect different brain patterns and be used in applications such as sensory evaluation of food. However, considering the computational cost and efficiency, EEG data with many channels has to face the critical issue of channel selection. This paper proposed a channel selection method called class activation mapping with attention (CAM-Attention). The CAM-Attention method combined a convolutional neural network with channel and spatial attention (CNN-CSA) model with a gradient-weighted class activation mapping (Grad-CAM) model. The CNN-CSA model exploited key features in EEG data by attention mechanism, and the Grad-CAM model effectively realized the visualization of feature regions. Then, channel selection was effectively implemented based on feature regions. Finally, the CAM-Attention method reduced the computational burden of taste EEG recognition and effectively distinguished the four tastes. In short, it has excellent recognition performance and provides effective technical support for taste sensory evaluation.
Achieving low duty cycle operation in low-power wireless networks in urban environments is complicated by the complex and variable dynamics of external interference and fading. We explore the use of reinforcement learning for achieving low power consumption for the task of optimal selection of channels. The learning relies on a hybrid of passive channel sampling for dealing with external interference and active channel sampling for dealing with fading. Our solution, Passive-Active Multi-armed bandit for LoRa (PAMLR, pronounced "Pamela"), balances the two types of samples to achieve energy-efficient channel selection: active channel measurements are tuned to an appropriately low level to update noise thresholds, and to compensate passive channel measurements are tuned to an appropriately high level for selecting the top-most channels from channel exploration using the noise thresholds. The rates of both types of samples are adapted in response to channel dynamics. Based on extensive testing in multiple environments in different cities, we validate that PAMLR can maintain excellent communication quality, as demonstrated by a low SNR regret compared to the optimal channel allocation policy, while substantially minimizing the energy cost associated with channel measurements.
Brain-assisted speech enhancement (BASE) aims to extract the target speaker in complex multi-talker scenarios using electroencephalogram (EEG) signals as an assistive modality, as the auditory attention of the listener can be decoded from electroneurographic signals of the brain. This facilitates a potential integration of EEG electrodes with listening devices to improve the speech intelligibility of hearing-impaired listeners, which was shown by the recently-proposed BASEN model. As in general the multichannel EEG signals are highly correlated and some are even irrelevant to listening, blindly incorporating all EEG channels would lead to a high economic and computational cost. In this work, we therefore propose a geometry-constrained EEG channel selection approach for BASE. We design a new weighted multi-dilation temporal convolutional network (WDTCN) as the backbone to replace the Conv-TasNet in BASEN. Given a raw channel set that is defined by the electrode geometry for feasible integration, we then propose a geometry-constrained convolutional regularization selection (GC-ConvRS) module for WD-TCN to find an informative EEG subset. Experimental results on a public dataset show the superiority of the proposed WD-TCN over BASEN. The GC-ConvRS can further refine the useful EEG subset subject to the geometry constraint, resulting in a better trade-off between performance and integration cost.
Color Filter Arrays (CFA) are optical filters in digital cameras that capture specific color channels. Current commercial CFAs are hand-crafted patterns with different physical and application-specific considerations. This study proposes a binary CFA learning module based on hard thresholding with a deep learning-based demosaicing network in a joint architecture. Unlike most existing learnable CFAs that learn a channel from the whole color spectrum or linearly combine available digital colors, this method learns a binary channel selection, resulting in CFAs that are practical and physically implementable to digital cameras. The binary selection is based on adapting the hard thresholding operation into neural networks via a straight-through estimator, and therefore it is named HardMax. This paper includes the background on the CFA design problem, the description of the HardMax method, and the performance evaluation results. The evaluation of the proposed method includes tests for different demosaicing models, color configurations, filter sizes, and a comparison with existing methods in various reconstruction metrics. The proposed approach is tested with Kodak and BSDS500 datasets and provides higher reconstruction performance than hand-crafted or alternative learned binary filters.
Explanation for Multivariate Time Series Classification (MTSC) is an important topic that is under explored. There are very few quantitative evaluation methodologies and even fewer examples of actionable explanation, where the explanation methods are shown to objectively improve specific computational tasks on time series data. In this paper we focus on analyzing InterpretTime, a recent evaluation methodology for attribution methods applied to MTSC. We showcase some significant weaknesses of the original methodology and propose ideas to improve both its accuracy and efficiency. Unlike related work, we go beyond evaluation and also showcase the actionability of the produced explainer ranking, by using the best attribution methods for the task of channel selection in MTSC. We find that perturbation-based methods such as SHAP and Feature Ablation work well across a set of datasets, classifiers and tasks and outperform gradient-based methods. We apply the best ranked explainers to channel selection for MTSC and show significant data size reduction and improved classifier accuracy.
In end-to-end multi-channel speech enhancement, the traditional approach of designating one microphone signal as the reference for processing may not always yield optimal results. The limitation is particularly in scenarios with large distributed microphone arrays with varying speaker-to-microphone distances or compact, highly directional microphone arrays where speaker or microphone positions change over time. Current mask-based methods often fix the reference channel during training, which makes it not possible to adaptively select the reference channel for optimal performance. To address this problem, we introduce an adaptive approach for selecting the optimal reference channel. Our method leverages a multi-channel masking-based scheme, where multiple masked signals are combined to generate a single-channel output signal. This enhanced signal is then used for loss calculation, while the reference clean speech is adjusted based on the highest scale-invariant signal-to-distortion ratio (SI-SDR). The experimental results on the Spear challenge simulated dataset D4 demonstrate the superiority of our proposed method over the conventional approach of using a fixed reference channel with single-channel masking.
In this article, we explore the potential of using latent diffusion models, a family of powerful generative models, for the task of reconstructing naturalistic music from electroencephalogram (EEG) recordings. Unlike simpler music with limited timbres, such as MIDI-generated tunes or monophonic pieces, the focus here is on intricate music featuring a diverse array of instruments, voices, and effects, rich in harmonics and timbre. This study represents an initial foray into achieving general music reconstruction of high-quality using non-invasive EEG data, employing an end-to-end training approach directly on raw data without the need for manual pre-processing and channel selection. We train our models on the public NMED-T dataset and perform quantitative evaluation proposing neural embedding-based metrics. Our work contributes to the ongoing research in neural decoding and brain-computer interfaces, offering insights into the feasibility of using EEG data for complex auditory information reconstruction.
This paper introduces DeltaKWS, to the best of our knowledge, the first $\Delta$RNN-enabled fine-grained temporal sparsity-aware KWS IC for voice-controlled devices. The 65 nm prototype chip features a number of techniques to enhance performance, area, and power efficiencies, specifically: 1) a bio-inspired delta-gated recurrent neural network ($\Delta$RNN) classifier leveraging temporal similarities between neighboring feature vectors extracted from input frames and network hidden states, eliminating unnecessary operations and memory accesses; 2) an IIR BPF-based FEx that leverages mixed-precision quantization, low-cost computing structure and channel selection; 3) a 24 kB 0.6 V near-$V_\text{TH}$ weight SRAM that achieves 6.6X lower read power than the foundry-provided SRAM. From chip measurement results, we show that the DeltaKWS achieves an 11/12-class GSCD accuracy of 90.5%/89.5% respectively and energy consumption of 36 nJ/decision in 65 nm CMOS process. At 87% temporal sparsity, computing latency and energy/inference are reduced by 2.4X/3.4X, respectively. The IIR BPF-based FEx, $\Delta$RNN accelerator, and 24 kB near-$V_\text{TH}$ SRAM blocks occupy 0.084 mm$^{2}$, 0.319 mm$^{2}$, and 0.381 mm$^{2}$ respectively (0.78 mm$^{2}$ in total).
Brain-computer interface systems and the recording of brain activity has garnered significant attention across a diverse spectrum of applications. EEG signals have emerged as a modality for recording neural electrical activity. Among the methodologies designed for feature extraction from EEG data, the method of RCSP has proven to be an approach, particularly in the context of MI tasks. RCSP exhibits efficacy in the discrimination and classification of EEG signals. In optimizing the performance of this method, our research extends to a comparative analysis with conventional CSP techniques, as well as optimized methodologies designed for similar applications. Notably, we employ the meta-heuristic multi-objective Strength Pareto Evolutionary Algorithm II (SPEA-II) as a pivotal component of our research paradigm. This is a state-of-the-art approach in the selection of an subset of channels from a multichannel EEG signal with MI tasks. Our main objective is to formulate an optimum channel selection strategy aimed at identifying the most pertinent subset of channels from the multi-dimensional electroencephalogram (EEG) signals. One of the primary objectives inherent to channel selection in the EEG signal analysis pertains to the reduction of the channel count, an approach that enhances user comfort when utilizing gel-based EEG electrodes. Additionally, within this research, we took benefit of ensemble learning models as a component of our decision-making. This technique serves to mitigate the challenges associated with overfitting, especially when confronted with an extensive array of potentially redundant EEG channels and data noise. Our findings not only affirm the performance of RCSP in MI-based BCI systems, but also underscore the significance of channel selection strategies and ensemble learning techniques in optimizing the performance of EEG signal classification.
Brain-computer interface (BCI) systems facilitate unique communication between humans and computers, benefiting severely disabled individuals. Despite decades of research, BCIs are not fully integrated into clinical and commercial settings. It's crucial to assess and explain BCI performance, offering clear explanations for potential users to avoid frustration when it doesn't work as expected. This work investigates the efficacy of different deep learning and Riemannian geometry-based classification models in the context of motor imagery (MI) based BCI using electroencephalography (EEG). We then propose an optimal transport theory-based approach using earth mover's distance (EMD) to quantify the comparison of the feature relevance map with the domain knowledge of neuroscience. For this, we utilized explainable AI (XAI) techniques for generating feature relevance in the spatial domain to identify important channels for model outcomes. Three state-of-the-art models are implemented - 1) Riemannian geometry-based classifier, 2) EEGNet, and 3) EEG Conformer, and the observed trend in the model's accuracy across different architectures on the dataset correlates with the proposed feature relevance metrics. The models with diverse architectures perform significantly better when trained on channels relevant to motor imagery than data-driven channel selection. This work focuses attention on the necessity for interpretability and incorporating metrics beyond accuracy, underscores the value of combining domain knowledge and quantifying model interpretations with data-driven approaches in creating reliable and robust Brain-Computer Interfaces (BCIs).