Chris Bartels

TL;DR: This discussion employs dynamic Bayesian networks (DBNs) and a DBN extension using the Graphical Model Toolkit's (GMTK's) basic template, a dynamic graphical model representation that is more suitable for speech and language systems.

TL;DR: This work reports on investigations into the use of articulatory features (AFs) for observation and pronunciation models in speech recognition, and investigates a model having multiple streams of AF states with soft synchrony constraints, for both audio-only and audio-visual recognition.

TL;DR: This work applies a novel data selection technique based on constrained submodular function maximization to subselecting a large set of acoustic data to train automatic speech recognition (ASR) systems and shows that training data can be reduced significantly, and that the technique outperforms both random selection and a previously proposed selection method utilizing comparable resources.

TL;DR: A model based on dynamic Bayesian networks (DBN) to integrate information from multiple audio and visual streams and an absolute improvement of about 4% in word accuracy in the -4 to 10db average case when making use of two audio and one video streams is indicated.

TL;DR: Voices Obscured In Complex Environmental Settings (VOICES) as mentioned in this paper is a large-scale dataset of speech recorded by far-field microphones in noisy room conditions, where audio was recorded in furnished rooms with background noise played in conjunction with foreground speech selected from the LibriSpeech corpus.