Concordia University in Montreal, Canada is currently accepting applications for postdoctoral fellowships valued at $50,000 per year (plus benefits) for each of two years, in the following discipline:

Research program title:

Deep Learning Assisted Approaches to Enhancement, Separation and Reconstruction of Speech and Audio Signals

Supervisor(s):

Wei-Ping Zhu

Reference Number:

8012

Provide a link to one journal/publication that you think would be a suitable forum for advertising the postdoctoral fellowship opportunity:

Y. Ji, W.-P. Zhu and B. Champagne, “Speech enhancement based on dictionary learning and low-rank matrix decomposition,” IEEE Access, vol. 7, pp. 4936-4947, January 2019.

Program description (up to 200 words):

Over the past decade, speech and audio processing technology has largely advanced and facilitated voice communication and daily life by means of smart phones, personal hand-held devices, hearing aids, home smart devices, etc. As enabling technology, speech enhancement and audio signal recovery from severely degraded observations has become increasingly important for various applications such as automatic speech recognition (ASR), virtual reality and robotics. However, speech and audio signal processing researchers have been challenged by the sophisticated and adverse environment where the audio and speech signals are captured. A typical real-life complex acoustic environment example is known as ‘cocktail-party’ where multiple speakers may interfere with each other in the presence of environmental music and ambient noise possibly added by room reverberation. Such severe and comprehensive degradations make the extraction of the desired audio/speech signal extremely difficult. As such, it is imperative and very important to develop advanced audio signal processing techniques that can mimic the complicated human auditory systems to detect and enhance the desired speech from interfering and/or mixed signals by suppressing the environmental noise, alleviating reverberant components and separating the source signals This project aims to develop low-complexity deep learning aided speech and audio signal processing techniques that can handle very complicated adverse acoustic conditions. The objective is to separate and reconstruct the desired speech and audio signals with high accuracy and low complexity and latency.

Academic qualifications required:

PhD in Electrical Engineering with knowledge and research experience in the fields of source separation, speech enhancement and audio signal processing assisted by deep learning.

Eligibility requirements:

• Applicants must not currently hold a postdoctoral appointment at Concordia
• Priority will be given to postdoctoral fellows who have obtained their PhD from another university although in exceptional cases Concordia graduates may be considered
• Applicants must adhere to the postdoctoral fellow eligibility criteria outlined in Concordia University’s Postdoctoral Policy

Timeline and Application Process:

The Postdoctoral Fellow must start his/her appointment by March 01, 2021.

Submission process:

• All documents must be submitted to Niyusha Samadi at (Horizon.Postdocs@concordia.ca)
• Please include the reference number with your application

Application checklist:

• One to three (1-3) page research statement demonstrating fit with the program described above
• Current curriculum vitae demonstrating research excellence and a capacity for leadership in the domain (maximum 5 pages).
• Two letters of reference from academic supervisors or current employers to be sent via e-mail directly to: Niyusha Samadi at (Horizon.Postdocs@concordia.ca)

Concordia University is a vibrant research and teaching environment, with state-of-the-art research facilities and many research centers. Concordia is located in Montreal, Canada, a diverse and creative city, often ranked as offering one of the best quality of living experiences in North America.