English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

This paper presents an optimized methodology to design and deploy SpeechEnhancement (SE) algorithms based on Recurrent Neural Networks (RNNs) on astate-of-the-art MicroController Unit (MCU), with 1+8 general-purpose RISC-Vcores. To achieve low-latency execution, we propose an optimized softwarepipeline interleaving parallel computation of LSTM or GRU recurrent blocks,featuring vectorized 8-bit integer (INT8) and 16-bit floating-point (FP16)compute units, with manually-managed memory transfers of model parameters. Toensure minimal accuracy degradation with respect to the full-precision models,we propose a novel FP16-INT8 Mixed-Precision Post-Training Quantization (PTQ)scheme that compresses the recurrent layers to 8-bit while the bit precision ofremaining layers is kept to FP16. Experiments are conducted on multiple LSTMand GRU based SE models trained on the Valentini dataset, featuring up to 1.24Mparameters. Thanks to the proposed approaches, we speed-up the computation byup to 4x with respect to the lossless FP16 baselines. Differently from auniform 8-bit quantization that degrades the PESQ score by 0.3 on average, theMixed-Precision PTQ scheme leads to a low-degradation of only 0.06, whileachieving a 1.4-1.7x memory saving. Thanks to this compression, we cut thepower cost of the external memory by fitting the large models on the limitedon-chip non-volatile memory and we gain a MCU power saving of up to 2.5x byreducing the supply voltage from 0.8V to 0.65V while still matching thereal-time constraints. Our design results 10x more energy efficient thanstate-of-the-art SE solutions deployed on single-core MCUs that make use ofsmaller models and quantization-aware training.

Accelerating RNN-based Speech Enhancement on a Multi-Core MCU with Mixed FP16-INT8 Post-Training Quantization