2022/11/01 by Xingchen Song, Di Wu, Song, Xingchen +15 · 2 citations
Engineering · Social Sciences · #Advanced Chemical Sensor Technologies #Advanced Computing and Algorithms #Audio and Speech Processing (eess.AS) #Computation and Language (cs.CL) #FOS: Computer and information sciences #FOS: Electrical engineering #I.2.7 #Sound (cs.SD) #Underwater Vehicles and Communication Systems #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2211.00522
openalex publication_date 2022/11/01 · openalex created_date 2022/11/07 · openalex updated_date 2026/07/28
In this paper, we present TrimTail, a simple but effective emission regularization method to improve the latency of streaming ASR models. The core idea of TrimTail is to apply length penalty (i.e., by trimming trailing frames, see Fig. 1-(b)) directly on the spectrogram of input utterances, which does not require any alignment. We demonstrate that TrimTail is computationally cheap and can be applied online and optimized with any training loss or any model architecture on any dataset without any extra effort by applying it on various end-to-end streaming ASR networks either trained with CTC loss [1] or Transducer loss [2]. We achieve 100 ∼ 200ms latency reduction with equal or even better accuracy on both Aishell-1 and Librispeech. Moreover, by using TrimTail, we can achieve a 400ms algorithmic improvement of User Sensitive Delay (USD) with an accuracy loss of less than 0.2.