Sorry, you need to enable JavaScript to visit this website.

In speaker-aware training, a speaker embedding is appended to DNN input features. This allows the DNN to effectively learn representations, which are robust to speaker variability.
We apply speaker-aware training to attention-based end- to-end speech recognition. We show that it can improve over a purely end-to-end baseline. We also propose speaker-aware training as a viable method to leverage untranscribed, speaker annotated data.

Blood pressure (BP) is a vital sign of the human body and an important parameter for early detection of cardiovascular diseases. It is usually measured using cuff-based devices or monitored invasively in critically-ill patients. This paper presents two techniques that enable continuous and noninvasive cuff-less BP estimation using photoplethysmography (PPG) signals with Convolutional Neural Networks (CNNs). The first technique is calibration-free.

In supervised machine learning, the assumption that training data is labelled correctly is not always satisfied. In this paper, we investigate an instance of labelling error for classification tasks in which the dataset is corrupted with out-of-distribution (OOD) instances: data that does not belong to any of the target classes, but is labelled as such. We show that detecting and relabelling certain OOD instances, rather than discarding them, can have a positive effect on learning.

In this talk we present statistical signal processing methodologies on a real-world application of using Commercial Microwave Links (CMLs) as opportunistic sensors for rain monitoring. We formulate an appropriate parameter estimation problem, taking advantage on the empirically evaluated statistics of the rain, and present a new methodology for rain estimation given only the quantized minimum and maximum radio signal level measurements, which are being logged regularly by the network management systems.

Multimodal image registration plays an important role in diagnosing and treating ophthalmologic diseases. In this paper, a deep learning framework for multimodal retinal image registration is proposed. The framework consists of a segmentation network, feature detection and description network, and an outlier rejection network, which focuses only on the globally coarse alignment step using the perspective transformation.

Recently, a hybrid analog-digital architecture has been proposed for multiuser MIMO transmission in the millimeter-wave spectrum using reflect-arrays. The architecture exhibits scalability and high energy-efficiency while keeping the transmitter cost-efficient. Inspired by this architecture, we design a secure multiuser hybrid analog-digital precoding scheme. This scheme utilizes the method of regularized least-squares to shape the downlink beamformers, such that the signal received via malicious terminals is effectively suppressed.

In this paper, we revisit the popular affinity matrix based on the anchor graph and point out that the spectral embedding obtained using symmetric normalized Laplacian is only a side view of the bipartite structure. Based on the analysis, we propose Fast Spectral Clustering based on the Random Walk Laplacian (FRWL) method to explicitly balance the popularity of anchors and the independence of data points, which is especially important for clustering of boundary points.