arXiv subjects

Clément Laroche

Publications and source records attributed to Clément Laroche.

11 recordsLinked to original sources

arXiv2026

Beyond Model Size: Redesigning LiSenNet for embedded speech enhancement

Deploying real-time speech enhancement on resource-constrained devices requires meeting strict latency, memory, and energy constraints. Microcontroller NPUs can accelerate neural inference under these constraints, but only through a restricted set of operators in static, integer-quantized graphs. Recent speech-enhancement networks have reduced parameter counts and MACs to levels nominally suitable for microcontrollers, but their operators and execution patterns often remain incompatible with restricted NPUs. We address this gap by redesigning LiSenNet, a 37k parameter sub-band dual-path model, for the STM32N6570-DK Neural-ART accelerator. We replace its recurrent bottleneck with convolutional frequency and temporal mixers, reformulate unsupported operations as static int8-compatible primitives, and use bounded decoder activations to preserve quality after quantization. On VoiceBank-DEMAND, the final NPU-compatible model matches or exceeds the recurrent LiSenNet baseline, reaching PESQ 3.08 versus 3.01 in FP32 and 3.01 versus 2.93 in int8. Deployed on a microcontroller, it processes each 16 ms input hop in 4.83 ms, corresponding to a real-time factor of 0.30. Stateless receptive-field recomputation is an order of magnitude slower at the same frame rate despite higher accelerator utilization. These results show that parameter count and operator compatibility, quantization range, and persistent streaming state must be co-designed to achieve efficient real-time speech enhancement on restricted NPUs.

eess.AS↗

arXiv2026

Does per-frame early exit pay? A compute-matched study of dynamic depth for on-device speech enhancement

Deep learning-based speech enhancement is increasingly deployed on-device in hearing aids, headsets, and earbuds. Most of these devices, however, can only accelerate static int8 graphs, so a depth-varying network must be implemented as several graphs, orchestrated by a policy. In this paper, we supervise every intermediate depth of one causal model, then we fine-tune its output heads to guarantee that deeper outputs are never worse than shallower ones. Using this training protocol, we can derive a family of static models that are more Pareto-efficient than their equivalently-sized counterparts trained from scratch on the same budget. Specifically, we achieve up to 0.11 higher PESQ for equivalent compute, and match the best PESQ at 30% less compute. We then quantize the models to int8 and measure the latency-quality frontier on an STM32N6 microcontroller. On VoiceBank-DEMAND, the dynamic enhancer lies on the same frontier as the static models, rather than trading quality for dynamic execution. Running the policy on the companion Cortex-M55 takes only 26 $μ$s per frame, while splitting the enhancer into separate NPU graphs adds 2.2% latency overhead. The cost of dynamic execution is therefore small.

eess.AS↗

arXiv2026

From Diet to Free Lunch: Estimating Auxiliary Signal Properties using Dynamic Pruning Masks in Speech Enhancement Networks

Speech Enhancement (SE) in audio devices is often supported by auxiliary modules for Voice Activity Detection (VAD), SNR estimation, or Acoustic Scene Classification to ensure robust context-aware behavior and seamless user experience. Just like SE, these tasks often employ deep learning; however, deploying additional models on-device is computationally impractical, whereas cloud-based inference would introduce additional latency and compromise privacy. Prior work on SE employed Dynamic Channel Pruning (DynCP) to reduce computation by adaptively disabling specific channels based on the current input. In this work, we investigate whether useful signal properties can be estimated from these internal pruning masks, thus removing the need for separate models. We show that simple, interpretable predictors achieve up to 93% accuracy on VAD, 84% on noise classification, and an R2 of 0.86 on F0 estimation. With binary masks, predictions reduce to weighted sums, inducing negligible overhead. Our contribution is twofold: on one hand, we examine the emergent behavior of DynCP models through the lens of downstream prediction tasks, to reveal what they are learning; on the other, we repurpose and re-propose DynCP as a holistic solution for efficient SE and simultaneous estimation of signal properties.

eess.AS↗

arXiv2025

Adaptive Slimming for Scalable and Efficient Speech Enhancement

Speech enhancement (SE) enables robust speech recognition, real-time communication, hearing aids, and other applications where speech quality is crucial. However, deploying such systems on resource-constrained devices involves choosing a static trade-off between performance and computational efficiency. In this paper, we introduce dynamic slimming to DEMUCS, a popular SE architecture, making it scalable and input-adaptive. Slimming lets the model operate at different utilization factors (UF), each corresponding to a different performance/efficiency trade-off, effectively mimicking multiple model sizes without the extra storage costs. In addition, a router subnet, trained end-to-end with the backbone, determines the optimal UF for the current input. Thus, the system saves resources by adaptively selecting smaller UFs when additional complexity is unnecessary. We show that our solution is Pareto-optimal against individual UFs, confirming the benefits of dynamic routing. When training the proposed dynamically-slimmable model to use 10% of its capacity on average, we obtain the same or better speech quality as the equivalent static 25% utilization while reducing MACs by 29%.

eess.AS↗

arXiv2024

Resource-Efficient Speech Quality Prediction through Quantization Aware Training and Binary Activation Maps

As speech processing systems in mobile and edge devices become more commonplace, the demand for unintrusive speech quality monitoring increases. Deep learning methods provide high-quality estimates of objective and subjective speech quality metrics. However, their significant computational requirements are often prohibitive on resource-constrained devices. To address this issue, we investigated binary activation maps (BAMs) for speech quality prediction on a convolutional architecture based on DNSMOS. We show that the binary activation model with quantization aware training matches the predictive performance of the baseline model. It further allows using other compression techniques. Combined with 8-bit weight quantization, our approach results in a 25-fold memory reduction during inference, while replacing almost all dot products with summations. Our findings show a path toward substantial resource savings by supporting mixed-precision binary multiplication in hard- and software.

cs.SD↗

arXiv2024

Towards a tailored mixed-precision sub-8-bit quantization scheme for Gated Recurrent Units using Genetic Algorithms

Despite the recent advances in model compression techniques for deep neural networks, deploying such models on ultra-low-power embedded devices still proves challenging. In particular, quantization schemes for Gated Recurrent Units (GRU) are difficult to tune due to their dependence on an internal state, preventing them from fully benefiting from sub-8bit quantization. In this work, we propose a modular integer quantization scheme for GRUs where the bit width of each operator can be selected independently. We then employ Genetic Algorithms (GA) to explore the vast search space of possible bit widths, simultaneously optimising for model size and accuracy. We evaluate our methods on four different sequential tasks and demonstrate that mixed-precision solutions exceed homogeneous-precision ones in terms of Pareto efficiency. In our results, we achieve a model size reduction between 25% and 55% while maintaining an accuracy comparable with the 8-bit homogeneous equivalent.

cs.LG↗

arXiv2023

Dynamic nsNet2: Efficient Deep Noise Suppression with Early Exiting

Although deep learning has made strides in the field of deep noise suppression, leveraging deep architectures on resource-constrained devices still proved challenging. Therefore, we present an early-exiting model based on nsNet2 that provides several levels of accuracy and resource savings by halting computations at different stages. Moreover, we adapt the original architecture by splitting the information flow to take into account the injected dynamism. We show the trade-offs between performance and computational complexity based on established metrics.

cs.SD↗

arXiv2023

On Crowdsourcing-design with Comparison Category Rating for Evaluating Speech Enhancement Algorithms

Speech enhancement techniques improve the quality or the intelligibility of an audio signal by removing unwanted noise. It is used as preprocessing in numerous applications such as speech recognition, hearing aids, broadcasting and telephony. The evaluation of such algorithms often relies on reference-based objective metrics that are shown to correlate poorly with human perception. In order to evaluate audio quality as perceived by human observers it is thus fundamental to resort to subjective quality assessment. In this paper, a user evaluation based on crowdsourcing (subjective) and the Comparison Category Rating (CCR) method is compared against the DNSMOS, ViSQOL and 3QUEST (objective) metrics. The overall quality scores of three speech enhancement algorithms from real time communications (RTC) are used in the comparison using the P.808 toolkit. Results indicate that while the CCR scale allows participants to identify differences between processed and unprocessed audio samples, two groups of preferences emerge: some users rate positively by focusing on noise suppression processing, while others rate negatively by focusing mainly on speech quality. We further present results on the parameters, size considerations and speaker variations that are critical and should be considered when designing the CCR-based crowdsourcing evaluation.

eess.AS↗

arXiv2022

Pesticide concentration monitoring: investigating spatio-temporal patterns in left censored data

Monitoring pesticide concentration is very important for public authorities given the major concerns for environmental safety and the likelihood for increased public health risks. An important aspect of this process consists in locating abnormal signals, from a large amount of collected data. This kind of data is usually complex since it suffers from limits of quantification leading to left censored observations, and from the sampling procedure which is irregular in time and space across measuring stations. The present manuscript tackles precisely the issue of detecting spatio-temporal collective anomalies in pesticide concentration levels, and introduces a novel methodology for dealing with spatio-temporal heterogeneity. The latter combines a change-point detection procedure applied to the series of maximum daily values across all stations, and a clustering step aimed at a spatial segmentation of the stations. Limits of quantification are handled in the change-point procedure, by supposing an underlying left-censored parametric model, piece-wise stationary. Spatial segmentation takes into account the geographical conditions, and may be based on river network, wind directions, etc. Conditionally to the temporal segment and the spatial cluster, one may eventually analyse the data and identify contextual anomalies. The proposed procedure is illustrated in detail on a data set containing the prosulfocarb concentration levels in surface waters in Centre-Val de Loire region.

stat.AP↗

arXiv2020

An Implicit Representation of Swept Volumes based on Local Shapes and Movements

We introduce a new way to implicitly represent swept volumes in 3D. We first implicitize the base volume and then apply the time-dependent rigid transformation to build an implicit representation of the swept volume. This way, we build a fitting representation, where geometric details generated by the rigid transformation and details of the base volume itself are both taken into account. This representation relies on a partition of a boundary box of the swept volume according to portions of the space related either to parts of the base volume or to time span of the transformation.

cs.CG↗

arXiv2017

Extending a Morse function to a non-orientable $3-$manifold

Considering a solid 3-dimensional Klein bottle and a collaring of its boundary, can we extend a generic $C^\infty$ non-singular function defined on the collaring to the full solid Klein bottle without critical points? We give a condition on the Reeb graph of the given function that is necessary and sufficient for the existence of such a non-singular extension.

math.GT↗