# Extremely Low Footprint End-to-End ASR System for Smart Device

arXiv: https://arxiv.org/abs/2104.05784 · HTML: https://mlsanity.com/paper/2104.05784/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 6 Apr 2021.

## Built on by (most cited first, 7 of 7)

- [FunASR: A Fundamental End-to-End Speech Recognition Toolkit](https://mlsanity.com/paper/2305.11013/) (2023)
- [Omni-sparsity DNN: Fast Sparsity Optimization for On-Device Streaming E2E ASR via Supernet](https://mlsanity.com/paper/2110.08352/) (2021)
- [ResidualTransformer: Residual Low-Rank Learning with Weight-Sharing for Transformer Layers](https://mlsanity.com/paper/2310.02489/) (2023)
- [Dynamic ASR Pathways: An Adaptive Masking Approach Towards Efficient Pruning of A Multilingual ASR Model](https://mlsanity.com/paper/2309.13018/) (2023)
- [Distilling Multi-Level X-vector Knowledge for Small-footprint Speaker Verification](https://mlsanity.com/paper/2303.01125/) (2023)
- [Beyond Universal Transformer: block reusing with adaptor in Transformer for automatic speech recognition](https://mlsanity.com/paper/2303.13072/) (2023)
- [Diagnostic-Driven Layer-Wise Compensation for Post-Training Quantization of Encoder-Decoder ASR Models](https://mlsanity.com/paper/2601.02455/) (2026)

## Abstract (arXiv)

Recently, end-to-end (E2E) speech recognition has become popular, since it can integrate the acoustic, pronunciation and language models into a single neural network, which outperforms conventional models. Among E2E approaches, attention-based models, e.g. Transformer, have emerged as being superior. Such models have opened the door to deployment of ASR on smart devices, however they still suffer from requiring a large number of model parameters. We propose an extremely low footprint E2E ASR system for smart devices, to achieve the goal of satisfying resource constraints without sacrificing recognition accuracy. We design cross-layer weight sharing to improve parameter efficiency and further exploit model compression methods including sparsification and quantization, to reduce memory storage and boost decoding efficiency. We evaluate our approaches on the public AISHELL-1 and AISHELL-2 benchmarks. On the AISHELL-2 task, the proposed method achieves more than 10x compression (model size reduces from 248 to 24MB), at the cost of only minor performance loss (CER reduces from 6.49% to 6.92%).