Audio-visual speech corpus for evaluating speech recognition performance in noisy environments

点击翻译按钮获取中文摘要

Audio-visual speech data are collected in a silent room and a vehicle for developing an audio-visual speech corpus which is used for evaluating speech recognition performance in noisy environments, especially in in-car environments. Acoustic noise and gamma values of images are used for simulating in-car environments over the recorded data in the silent room. Baseline audio and visual features and an integration method are calibrated in some experimental evaluations. The corpus will be open to the public along with database manuals for research purposes.