Muhammed Shifas PV

Mathematics was my favourite subject in my old school days. Every time when I solve some problems I was wondered where it would be useful. Now I can see it on my day to day research, as being part of this ENRICH network. My research is mainly focused on developing signal processing model for speech intelligibility enhancement . As today's signal processing has been grown up to learn from the data (Machine Learning) we develop advanced Neural architectures for speech enhancement task. I am with Speech Signal Processing Lab (SSPL), University of Crete ( UoC), Greece from March. 2017 onward, supervised by Prof. Yannis Stylianou.

On one side, I work on developing advanced neural network models like WaveNet and FFTNet to improve listening comfort in noise, generally known as near end listening enhancement. This was achieved by modifying the speech characteristics, both in time and frequency domain. Our recently proposed wSSDRC model for this task can be seen on the webpage. In parallel, I also have suggested few novel feature extraction approaches for speech denoising. This new models, SE-FFTNet and gruCNN-SE for speech denoising, are displayed on the webpage for details. I collaborate with General Hospital, Chania, Greece, to evaluate our models. I was a visiting researcher at Sonova AG, Switzerland, where I contributed on developing shallow neural model for hearing aid devices. I am also interested on evaluating developed models on Listening Effort scale, specifically through EEG measures. For this I visited the Fraunhofer IDMT, Oldenburg, Germany in 2019.

I did my Bachelor degree (B.Tech) at University of Calicut, India, in the years 2009-2013. After, I completed my Master degree at the National Institute of Technology Calicut (NITC), India in 2014-2016. In my master thesis, I worked on developing modulation domain echo cancellation techniques directing to mobile and hand free devices. The thesis was superviosed by Prof. Sathidevi P.S, at the department of ECE, NITC.