Effects of Age and Uncertainty on the Visual Speech Benefit in Noise

Purpose: Listeners understand significantly more speech in noise when the talker's face can be seen (visual speech) in comparison to an auditory-only baseline (a visual speech benefit). This study investigated whether the visual speech benefit is reduced when the correspondence between auditory and visual speech is uncertain and whether any reduction is affected by listener age (older vs. younger) and how severe the auditory signal is masked. Method: Older and younger adults completed a speech recognition in noise task that included an auditory-only condition and four auditory–visual (AV) conditions in which one, two, four, or six silent talking face videos were presented. One face always matched the auditory signal; the other face(s) did not. Auditory speech was presented in noise at −6 and −1 dB signal-to-noise ratio (SNR). Results: When the SNR was −6 dB, for both age groups, the standard-sized visual speech benefit reduced as more talking faces were presented. When the SNR was −1 dB, younger adults received the standard-sized visual speech benefit even when two talking faces were presented, whereas older adults did not. Conclusions: The size of the visual speech benefit obtained by older adults was always smaller when AV correspondence was uncertain; this was not the case for younger adults. Difficulty establishing AV correspondence may be a factor that limits older adults' speech recognition in noisy AV environments. Supplemental Material https://doi.org/10.23641/asha.16879549

Efficient End-to-End Sentence-Level Lipreading with Temporal Convolutional Networks

Lipreading aims to recognize sentences being spoken by a talking face. In recent years, the lipreading method has achieved a high level of accuracy on large datasets and made breakthrough progress. However, lipreading is still far from being solved, and existing methods tend to have high error rates on the wild data and have the defects of disappearing training gradient and slow convergence. To overcome these problems, we proposed an efficient end-to-end sentence-level lipreading model, using an encoder based on a 3D convolutional network, ResNet50, Temporal Convolutional Network (TCN), and a CTC objective function as the decoder. More importantly, the proposed architecture incorporates TCN as a feature learner to decode feature. It can partly eliminate the defects of RNN (LSTM, GRU) gradient disappearance and insufficient performance, and this yields notable performance improvement as well as faster convergence. Experiments show that the training and convergence speed are 50% faster than the state-of-the-art method, and improved accuracy by 2.4% on the GRID dataset.