Hot-Fixing Wake Word Recognition for End-to-End ASR Via Neural Model Reprogramming
Lists
Only you see this
Abstract
This paper proposes two novel variants of neural reprogramming to enhance wake word recognition in streaming end-to-end ASR models without updating model weights. The first, "trigger-frame reprogramming", prepends the input speech feature sequence with the learned trigger-frames of the target wake word to adjust ASR model’s hidden states for improved wake word recognition. The second, "predictor-state initialization", trains only the initial state vectors (cell and hidden states) of the LSTMs in the prediction network. When applying to a baseline LibriSpeech Emformer RNN-T model with a 98% wake word verification false rejection rate (FRR) on unseen wake words, the proposed approaches achieve 76% and 97% relative FRR reductions with no increase on false acceptance rate. In-depth characteristic analyses of the proposed approaches are also conducted to provide deeper insights. These approaches offer an effective hot-fixing methods to improve wake word recognition performance in deployed production ASR models without the need for model updates.
BibTeX
@inproceedings{ku2024hot,
author = {Pin-Jui Ku and I-Fan Chen and Chao-Han Huck Yang and Anirudh Raju and Pranav Dheram and Pegah Ghahremani and Brian King and Jing Liu and Roger Ren and Phani Sankar Nidadavolu},
title = {{Hot-Fixing Wake Word Recognition for End-to-End ASR Via Neural Model Reprogramming}},
booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
year = {2024},
doi = {10.1109/icassp48485.2024.10448320},
}Loading related content...