......
Proposed a use of approximate circuits in non mission critical communication systems to gain PPA (Power
performance area) improvement in communication hardware.
Developed a methodology to implement best possible hardware.
Showcased the methodology on fundamental OFDM communication system with marginal loss of accuracy.
Proposed a methodology to improve hardware efficiency of bio medical signal processing using approximate
circuits.
Using the methodology, developed a hardware with significant PPA improvements for Pan-Tompkins Algorithm which is a filtering technique for ECG signals.
Automated RTL generation of LightGBM which is a gradient boosting framework that uses tree based learning algorithm.
This is a Python based library which performs all the required steps including Verilog generation and verification
form trained LightGBM classification model.
On FPGA, around 700 to 1000X speedup can be achieved for different classification data-sets.
Developed a low-cost, user-friendly, IoT enabled, head-phone sized tympanometer that detects middle ear problems.
The purchase of DHVANIK for such customers is analogous to a diabetic patient having a blood glucose monitor.
It is a dead reckoning based device for indoor navigation with energy-efficient, cost-effective, reliable hardware.
The device generates audio and haptic feedback for navigation especially for the visually impaired.