Hybrid Transformer/CTC Networks for Hardware Efficient Voice Triggering

Adya, Saurabh; Garg, Vineet; Sigtia, Siddharth; Simha, Pramod; Dhir, Chandra

doi:10.48550/arxiv.2008.02323

Cited by 1 publication

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Conventional KWS system has been developed maturely, including large vocabulary continuous speech recognition (LVCSR) based lattice search [1,2,3], hidden Markov model (HMM) based keyword-filler method [4,5,6], discriminative models based on large-margin formulation or recurrent networks and query-by-example (QbyE) based template matching approaches [7,8,9,10,11]. Recently, with the development of deep learning and its successful applications, deep KWS frameworks have been introduced [12,13,14,15,16,17,18]. In the deep KWS family, an acoustic model is trained to predict the sub-word of keyword and a posterior handling method is followed to generate a confidence score of the whole keyword.…”

Section: Introductionmentioning

confidence: 99%

IEEE SLT 2021 Alpha-mini Speech Challenge: Open Datasets, Tracks, Rules and Baselines

Fu¹,

Yao²,

He³

et al. 2020

Preprint

View full text Add to dashboard Cite

The IEEE Spoken Language Technology Workshop (SLT) 2021 Alpha-mini Speech Challenge (ASC) is intended to improve research on keyword spotting (KWS) and sound source location (SSL) on humanoid robots. Many publications report significant improvements in deep learning based KWS and SSL on open source datasets in recent years. For deep learning model training, it is necessary to expand the data coverage to improve the model robustness. Thus, simulating multi-channel noisy and reverberant data from single-channel speech, noise, echo and room impulsive response (RIR) is widely adopted. However, this approach may generate mismatch between simulated data and recorded data in real application scenarios, especially echo data. In this challenge, we open source a sizable speech, keyword, echo and noise corpus for promoting data-driven methods, particularly deep-learning approaches on KWS and SSL. We also choose Alpha-mini, a humanoid robot produced by UBTECH equipped with a built-in four-microphone array on its head, to record development and evaluation sets under the actual Alpha-mini robot application scenario, including environmental noise as well as echo and mechanical noise generated by the robot itself for model evaluation. Furthermore, we illustrate the rules, evaluation methods and baselines for researchers to quickly assess their achievements and optimize their models.

show abstract