Wearable devices like smart glasses are approaching the compute capability to seamlessly generate real-time closed captions for live conversations. We build on our recently introduced directional Automatic Speech Recognition (ASR) for smart glasses that have microphone arrays, which fuses multi-channel ASR with serialized output training, for wearer/conversation-partner disambiguation as well as suppression of cross-talk speech from non-target directions and noise. When ASR work is part of a broader system-development process, one may be faced with changes to microphone geometries as system development progresses. This paper aims to make multi-channel ASR insensitive to limited variations of microphone-array geometry. We show that a model trained on multiple similar geometries is largely agnostic and generalizes well to new geometries, as long as they are not too different. Furthermore, training the model this way improves accuracy for seen geometries by 15 to 28\% relative. Lastly, we refine the beamforming by a novel Non-Linearly Constrained Minimum Variance criterion.
Modern smart glasses leverage advanced audio sensing and machine learning technologies to offer real-time transcribing and captioning services, considerably enriching human experiences in daily communications. However, such systems frequently encounter challenges related to environmental noises, resulting in degradation to speech recognition and speaker change detection. To improve voice quality, this work investigates directional source separation using the multi-microphone array. We first explore multiple beamformers to assist source separation modeling by strengthening the directional properties of speech signals. In addition to relying on predetermined beamformers, we investigate neural beamforming in multi-channel source separation, demonstrating that automatic learning directional characteristics effectively improves separation quality. We further compare the ASR performance leveraging separated outputs to noisy inputs. Our results show that directional source separation benefits ASR for the wearer but not for the conversation partner. Lastly, we perform the joint training of the directional source separation and ASR model, achieving the best overall ASR performance.
End-to-end (E2E) spoken language understanding (SLU) systems that generate a semantic parse from speech have become more promising recently. This approach uses a single model that utilizes audio and text representations from pre-trained speech recognition models (ASR), and outperforms traditional pipeline SLU systems in on-device streaming scenarios. However, E2E SLU systems still show weakness when text representation quality is low due to ASR transcription errors. To overcome this issue, we propose a novel E2E SLU system that enhances robustness to ASR errors by fusing audio and text representations based on the estimated modality confidence of ASR hypotheses. We introduce two novel techniques: 1) an effective method to encode the quality of ASR hypotheses and 2) an effective approach to integrate them into E2E SLU models. We show accuracy improvements on STOP dataset and share the analysis to demonstrate the effectiveness of our approach.
This paper studies audio-visual suppression for egocentric videos -- where the speaker is not captured in the video. Instead, potential noise sources are visible on screen with the camera emulating the off-screen speaker's view of the outside world. This setting is different from prior work in audio-visual speech enhancement that relies on lip and facial visuals. In this paper, we first demonstrate that egocentric visual information is helpful for noise suppression. We compare object recognition and action classification based visual feature extractors, and investigate methods to align audio and visual representations. Then, we examine different fusion strategies for the aligned features, and locations within the noise suppression model to incorporate visual information. Experiments demonstrate that visual features are most helpful when used to generate additive correction masks. Finally, in order to ensure that the visual features are discriminative with respect to different noise types, we introduce a multi-task learning framework that jointly optimizes audio-visual noise suppression and video based acoustic event detection. This proposed multi-task framework outperforms the audio only baseline on all metrics, including a 0.16 PESQ improvement. Extensive ablations reveal the improved performance of the proposed model with multiple active distractors, over all noise types and across different SNRs.
Cannulation for hemodialysis is the act of inserting a needle into a surgically created vascular access (e.g., an arteriovenous fistula) for the purpose of dialysis. The main risk associated with cannulation is infiltration, the puncture of the wall of the vascular access after entry, which can cause medical complications. Simulator-based training allows clinicians to gain cannulation experience without putting patients at risk. In this paper, we propose to use deep-learning-based techniques for detecting, based on video, whether the needle tip is in or has infiltrated the simulated fistula. Three categories of deep neural networks are investigated in this work: modified pre-trained models based on VGG-16 and ResNet-50, light convolutional neural networks (light CNNs), and convolutional recurrent neural networks (CRNNs). CRNNs consist of convolutional layers and a long short-term memory (LSTM) layer. A data set of cannulation experiments was collected and analyzed. The results show that both the light CNN and the CRNN achieve better performance than the pre-trained baseline models. The CRNN was implemented in real time on commodity hardware for use in the cannulation simulator, and the performance was verified. Deep-learning video analysis is a viable method for detecting needle state in a low cost cannulation simulator. Our data sets and code are released at https://github.com/axin233/DL_for_Needle_Detection_Cannulation
Multi-stage learning is an effective technique to invoke multiple deep-learning modules sequentially. This paper applies multi-stage learning to speech enhancement by using a multi-stage structure, where each stage comprises a self-attention (SA) block followed by stacks of temporal convolutional network (TCN) blocks with doubling dilation factors. Each stage generates a prediction that is refined in a subsequent stage. A fusion block is inserted at the input of later stages to re-inject original information. The resulting multi-stage speech enhancement system, in short, multi-stage SA-TCN, is compared with state-of-the-art deep-learning speech enhancement methods using the LibriSpeech and VCTK data sets. The multi-stage SA-TCN system's hyper-parameters are fine-tuned, and the impact of the SA block, the fusion block and the number of stages are determined. The use of a multi-stage SA-TCN system as a front-end for automatic speech recognition systems is investigated as well. It is shown that the multi-stage SA-TCN systems perform well relative to other state-of-the-art systems in terms of speech enhancement and speech recognition scores.