About US

EMST is a research and development laboratory focussing on all aspects of biomedical and speech processing areas. The group includes faculty, staff, research scholoars, project engineers, graduate and undergraduate students. The members of the group develop new methods and systems in this areas. The natural signals from human system like speech, fundus image, ECG, EMG, EEG and handwriting are the most challenging signals with rich information that are of interest to the signal processing and pattern recognition community. These signals are non-stationary in nature and provide a grand scope for developing different signal processing methods for automatic extraction of relevant information from them.

Some of the challenging issues that we focus include how to extract: speaker specific excitation information, robust features for children speech recognition, robust features for stressed speech processing, blood vessels, optic disc and lesions from fundus image, R-peak and QRS-complex from ECG, vowel and non-vowel like regions from speech and basic units from handwriting for automatic modelling. After extracting relevant information from the natural signals, the next step is to develop systems for automatic processing. Some of the systems that we aim to develop include speaker recognition system, stressed speech recognition system, children speech recognition system, speech synthesis system, fundus image processing system and multi dimensional ECG signal processing system.

We undertake sponsored and consultancy projects in all aspects of biomedical and speech processing areas. We also develop innovative, indegenous and low cost technology suitable for Indian scenario through our spin-off SpeecHWareNet.

Our broad fields of interest include the following:

  • Speech Signal Processing
  • Speech Enhancement
  • Speaker Recognition
  • Speech Synthesis
  • Speech Recognition
  • Biomedical Signal Processing
  • Medical Image Processing
  • Handwriting Data Processing

Our current topics of interest include the following:

  • Temporal and Spectral Processing Speech Enhancement Methods
  • Limited Data Speaker Recognition
  • Stressed Speaker Recognition
  • Childrens Speech Recognition
  • Modelling Speaker Information from LP residual
  • Stressed Speech Recognition
  • Fundus Image Processing
  • Multi Channnel ECG Processing
  • Emotional Speech Synthesis
  • Dynamic Prosody Modification
  • Neutral to Emotion Conversion
  • Bandwidth Expansion of Speech
  • Detection of Vowel and Non-Vowel Like Regions
  • Language, Dialect and Accent Identification
  • Handwriting Data Processing