Distilled Non-Semantic Speech Embeddings with Binary Neural Networks for Low-Resource Devices

Abstract

This work introduces BRILLsson, a novel binary neural network-based representation learning model for a broad range of non-semantic speech tasks. We train the model with knowledge distillation from a large and real-valued TRILLsson model with only a fraction of the dataset used to train TRILLsson. The resulting BRILLsson models are only 2MB in size with a latency less than 8ms, making them suitable for deployment in low-resource devices such as wearables. We evaluate BRILLsson on eight benchmark tasks (including but not limited to spoken language identification, emotion recognition, health condition diagnosis, and keyword spotting), and demonstrate that our proposed ultra-light and low-latency models perform as well as large-scale models.

References in corpus (6)

• Leveraging Recent Advances in Deep Learning for Audio-Visual Emotion Recognition
• BigSSL: Exploring the Frontier of Large-Scale Semi-Supervised Learning for Automatic Speech Recognition
• Back to Simplicity: How to Train Accurate BNNs from Scratch?
• Vocalsound: A Dataset for Improving Human Vocal Sounds Recognition
• TRILLsson: Distilled Universal Paralinguistic Speech Representations
• FRILL: A Non-Semantic Speech Embedding for Mobile Devices