通过掩码自蒸馏缓解RNN-T过拟合，提升关键词检测性能。
Masked Self-distilled Transducer-based Keyword Spotting with Semi-autoregressive Decoding
- 用掩码自蒸馏训练，让模型不依赖预测网络。
- 半自回归解码在保持高精度的同时抑制过拟合。
- 适合语音关键词检测场景，尤其在复杂环境下表现更好。
基于RNN-T的关键词检测（KWS）采用自回归解码（AR）因其流式架构和优异性能受到关注。然而，RNN-T中预测网络结构简单，容易在挑战性场景下过拟合，导致性能下降。本文提出一种掩码自蒸馏（MSD）训练策略，避免RNN-T过度依赖预测网络，从而缓解过拟合。该方法支持完全掩码预测输出的非自回归（NAR）解码。此外，我们设计了一种半自回归（SAR）解码方式，融合AR与NAR的优势。在多个KWS数据集上的实验表明，MSD训练有效缓解过拟合；SAR解码在保持AR解码高性能的同时，受益于NAR解码的抗过拟合能力，取得卓越效果。
原文摘要 · Abstract (English)
RNN-T-based keyword spotting (KWS) with autoregressive decoding~(AR) has gained attention due to its streaming architecture and superior performance. However, the simplicity of the prediction network in RNN-T poses an overfitting issue, especially under challenging scenarios, resulting in degraded performance. In this paper, we propose a masked self-distillation (MSD) training strategy that avoids RNN-Ts overly relying on prediction networks to alleviate overfitting. Such training enables masked non-autoregressive (NAR) decoding, which fully masks the RNN-T predictor output during KWS decoding. In addition, we propose a semi-autoregressive (SAR) decoding approach to integrate the advantages of AR and NAR decoding. Our experiments across multiple KWS datasets demonstrate that MSD training effectively alleviates overfitting. The SAR decoding method preserves the superior performance of AR decoding while benefits from the overfitting suppression of NAR decoding, achieving excellent results.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。