Most of the prior studies in the spatial Direction of Arrival (DoA) domain focus on a single modality. However, humans use auditory and visual senses to detect the presence of sound sources. With this motivation, we propose to use neural networks with audio and visual signals for multi-speaker localization. The use of heterogeneous sensors can provide complementary information to overcome uni-modal challenges, such as noise, reverberation, illumination variations, and occlusions. We attempt to address these issues by introducing an adaptive weighting mechanism for audio-visual fusion. We also propose a novel video simulation method that generates visual features from noisy target 3D annotations that are synchronized with acoustic features. Experimental results confirm that audio-visual fusion consistently improves the performance of speaker DoA estimation, while the adaptive weighting mechanism shows clear benefits.
Speaker and utterance verification are two tasks that co-exist in text-dependent speaker verification (SV), where a phrase of the same lexical information is spoken during train and test sessions. The conventional approaches mostly verify the speaker and the utterance separately using two models. While there are studies on joint modeling of speaker and utterance, it is always desirable to have a common framework that performs both speaker and utterance verification to access the intended service. To this end, we propose a unified framework that deals with both objectives and the trade-off between the two. The unified framework is based on long short term memory network trained using both speaker and utterance information. We use Part I of RSR2015 database for the studies in this work. We show that the unified framework not only demonstrates competitive SV performance, but also provides a solution for a system to address different levels of security need.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.