A5083490412- Speech and Audio Processing
- Autophagy in Disease and Therapy
- Advanced Adaptive Filtering Techniques
- Parkinson's Disease Mechanisms and Treatments
- Ubiquitin and proteasome pathways
- Speech Recognition and Synthesis
- Indoor and Outdoor Localization Technologies
- Protein Degradation and Inhibitors
- Music and Audio Processing
- Integrated Circuits and Semiconductor Failure Analysis
- Advanced Data Compression Techniques
- Noise Effects and Management
- Underwater Vehicles and Communication Systems
- Hand Gesture Recognition Systems
- Gait Recognition and Analysis
- Advancements in Photolithography Techniques
- Mitochondrial Function and Pathology
- Advanced Image and Video Retrieval Techniques
- Human Pose and Action Recognition
- Neurological disorders and treatments
- Visual Attention and Saliency Detection
- Underwater Acoustics Research
- Gaze Tracking and Assistive Technology
- Cancer, Hypoxia, and Metabolism
- Microtubule and mitosis dynamics
Google (United States)
2023-2024
McGill University
2017-2022
Article2 May 2022Open Access Transparent process Structural basis for feedforward control in the PINK1/Parkin pathway Véronique Sauvé orcid.org/0000-0002-5981-4573 Department of Biochemistry and Centre de Recherche en Biologie Structurale, McGill University, Montreal, QC, Canada Contribution: Conceptualization, Data curation, Formal analysis, Investigation, Visualization, Methodology, Writing - original draft, review & editing Search more papers by this author George Sung...
We present StreamVC, a streaming voice conversion solution that preserves the content and prosody of any source speech while matching timbre from target speech. Unlike previous approaches, StreamVC produces resulting waveform at low latency input signal even on mobile platform, making it applicable to real-time communication scenarios like calls video conferencing, addressing use cases such as anonymization in these scenarios. Our design leverages architecture training strategy SoundStream...
We present an on-device real-time hand gesture recognition (HGR) system, which detects a set of predefined static gestures from single RGB camera. The system consists two parts: skeleton tracker and classifier. use MediaPipe Hands as the basis tracker, improve keypoint accuracy, add estimation 3D keypoints in world metric space. create different classifiers, one based on heuristics other using neural networks (NN).
This study was designed to investigate the amount of directionality among three directional aids different brands and compare speech discrimination ability hearing-impaired persons using nondirectional hearing in competing background noise. For each aid under study, full-on gain curves were obtained at 45, 135, 225, 315 degree azimuths an anechoic chamber. Thirty-two users included. Discrimination assessed, commercially available W-22 disc recordings. Findings indicate varies aids. The that...
We present StreamVC, a streaming voice conversion solution that preserves the content and prosody of any source speech while matching timbre from target speech. Unlike previous approaches, StreamVC produces resulting waveform at low latency input signal even on mobile platform, making it applicable to real-time communication scenarios like calls video conferencing, addressing use cases such as anonymization in these scenarios. Our design leverages architecture training strategy SoundStream...
ABSTRACT PINK1 and parkin constitute a mitochondrial quality control system mutated in Parkinson’s disease. PINK1, kinase, phosphorylates ubiquitin to recruit parkin, an E3 ligase, mitochondria. controls both localization activity through phosphorylation of the ubiquitin-like (Ubl) domain parkin. Here, we observe that phospho-ubiquitin can bind two distinct sites on high affinity site RING1 localization, low RING0 releases autoinhibition. Surprisingly, NMR titrations vinyl sulfone assays...
Monitoring detectivity of the wafer edge, bevel and apex - areas beyond pattern is becoming increasingly important in yield enhancement efforts high-end fabs. In this paper we present a methodology for root cause analysis edge defects, based on inline SEM review EDX-based material analysis.
High quality speech capture has been widely studied for both voice communication and human computer interface reasons. To improve the performance, we can often find multi-microphone enhancement techniques deployed on various devices. Multi-microphone problem is decomposed into two decoupled steps: a beamformer that provides spatial filtering single-channel model cleans up output. In this work, propose solution takes raw microphone outputs as input an ML model. We devise simple yet effective...
We propose a neural network model that can separate target speech sources from interfering at different angular regions using two microphones. The is trained with simulated room impulse responses (RIRs) omni-directional microphones without needing to collect real RIRs. By relying on specific and multiple simulations, the utilizes consistent time difference of arrival (TDOA) cues, or what we call delay contrast, interference while remaining robust in various reverberation environments....
We propose a neural network model that can separate target speech sources from interfering at different angular regions using two microphones. The is trained with simulated room impulse responses (RIRs) omnidirectional microphones without needing to collect real RIRs. By relying on specific and multiple simulations, the utilizes consistent time difference of arrival (TDOA) cues, or what we call delay contrast, interference while remaining robust in various reverberation environments....
We present a novel self-contained camera-projector tabletop system with lamp form-factor that brings digital intelligence to our tables. propose real-time, on-device, learning-based touch detection algorithm makes any interactive. The top-down configuration and method robust the presence of clutter, main limitation existing systems. Our research prototype enables set experiences combine hand interactions objects on table. A video can be found at https://youtu.be/hElC_c25Fg8.
We introduce an efficient video segmentation system for resource-limited edge devices leveraging heterogeneous compute. Specifically, we design network models by searching across multiple dimensions of specifications the neural architectures and operations on top already light-weight backbones, targeting commercially available inference engines. further analyze optimize data flows in our systems CPU, GPU NPU. Our approach has empirically factored well into real-time AR system, enabling...
High quality speech capture has been widely studied for both voice communication and human computer interface reasons. To improve the performance, we can often find multi-microphone enhancement techniques deployed on various devices. Multi-microphone problem is decomposed into two decoupled steps: a beamformer that provides spatial filtering single-channel model cleans up output. In this work, propose solution takes raw microphone outputs as input an ML model. We devise simple yet effective...
Background: The E3 ubiquitin ligases can be subdivided into four distinct types (RING, HECT, U-box, and RBR type) based on their domain architecture transfer mechanism. Recent structures of different have been solved showing enzymes in autoinhibited state. only exception is HOIP/ HOIL-1L which was recently its “active” conformation. This review discusses the structural functional characteristics three members ligase family: Parkin, HOIP/HOIL-1L, HHARI. Methods: Searches were performed using...
The Parkinson disease associated proteins, Parkin and PINK1, together comprise a mitochondrial quality control system that promotes neuronal survival through autophagy of damaged mitochondria. In the pathway, PINK1 acts as sensor depolarized mitochondria phosphorylates ubiquitin to recruit activate on outer membrane. ubiquitinates which leads removal mitophagy. We have carried out structural functional studies understand its regulation mechanism activation 1 – 3 . exhibits low basal activity...