Welcome to SLangLab! Speech and Language Research Lab at NEU
No speaker left behind: advancing speech technology for disordered speech
People with motor speech disorders such as dysarthria, which impacts a majority of stroke patients, often have difficulty being understood by their communication partners. This often leads to communication difficulties and breakdowns. Enabling speakers with motor impairments to communicate
with greater clarity and efficacy helps ensure they can participate actively in society, feeling included and contributing to their full potential. Our research aims to convert low-intelligibility output from dysarthric speakers into accurate text and a clear audio message using automatic speech recognition (ASR) and voice conversion (VC).
Inclusive speech processing
Machine perception for atypical speech patterns.
Voice collector
A data collection platform for atypical speech patterns.
Aligning atypical and typical speech
Using neural audio representations and dynamic time warping to compare atypical and typical speech
Lab / Research Overview
Machine Learning
Developing machine learning algorithms that recognize speech patterns of people with dysarthria and help facilitate effective communication will help improve the quality of life of thousands of people each year.
Natural Language Processing
Our work looks at applications from anomaly prediction, to working with large language models for various natural language processing applications.
Speech Recognition
The use of foundation models for effective transfer learning has fundamentally changed machine learning system development. The project aims to investigate foundation models for voice conversion with speech recognition-based models.
Contact us
