It Takes Few to TANGO: A Quantized Distributed Model for Binaural Speech Enhancement
2026-07-09 • Sound
Sound
AI summaryⓘ
The authors studied how to make a complex speech enhancement system, called TANGO, use less computing power and memory by using low-precision calculations. They tested ways to reduce the accuracy of neural network parts and found that even though some parts became less precise, the system’s later processing fixed most errors. Using this insight, they created a simpler version named MN-TANGO that is smaller and faster but still works about as well. Their updated system uses techniques like INT8 quantization to run efficiently on devices with limited resources.
neural networksspeech enhancementquantizationpost-training quantizationquantization-aware trainingspatial filteringbinaural processingINT8 quantizationcomputational complexitymodel compression
Authors
Zahra Benslimane, Pierre Chouteau, Martyna Poreba, Fabrice Auzanneau, Michal Szczepanski, Fabian Chersi, Romain Serizel
Abstract
Neural network-based multichannel speech enhancement systems achieve strong enhancement performance, but their computational and memory requirements limit deployment on resource-constrained devices. This paper investigates low-precision inference for TANGO, a hybrid distributed binaural speech enhancement system combining neural mask estimation with spatial filtering. We evaluate post-training quantization and quantization-aware training for the neural components, and analyze how quantization errors in the mask estimators propagate through the downstream spatial filtering stage. Our analysis shows that, although quantization degrades intermediate mask estimates, the spatial filtering stage compensates for most quantization-induced errors. Leveraging this robustness, we simplify TANGO into MN-TANGO, reducing both model size and computational complexity while maintaining comparable final performance. By combining INT8 weight-and-activation quantization with ERB compression and grouped recurrent layers, the most compact MN-TANGO reaches 4.65 MMAC/s and 0.177 MB.