Bhavesh Sonwani

Multi-LTE interface-based Smart Ambulance System using Raspberry Pi

A multi-access solution for streaming ultra-high-definition video in the uplink from mobile systems. This solution intelligently divides and distributes video traffic across multiple network interfaces, incorporating inter-stream coding to ensure reliable transmission. The primary goal is to enhance and manage end-to-end delay and jitter performance in multi-access scenarios. A standardized set of APIs is proposed to access smart-ambulance services, providing live data for improved information utilization. This empowers hospital systems to be highly responsive and adaptable during critical situations in the pre-hospital and post-hospital phases of patient care. Note: The visibility has been currently changed to private. Please ping me for access.

Smart Radio Environment

A prototype implementation of a Smart Radio Environment (SRE) based communication system. A reflecting plate, a reconfigurable intelligent surface (RIS), is placed between the transmitter and receiver to build the SRE. With this setup, the received power improvement of around 20 dBm has been achieved if RIS is deployed. Also, an algorithm at GRC finds the beamforming angle where the received power is maximized.

Self-Supervised Representation Learning of Shapes: CL on 3D PointCloud Dataset using PyTorch & PyTorchGeometric

PyTorch geometric implementation of DynamicEdgeConv. It uses the parameter k to detect the nearest neighbours which form a subgraph. Used InfoNCE / NT-Xent Loss implemented in pytorch metric learning library. Temperature allows us to balance the similarity measure (making it more peaked) with typical values around 0.1 / 0.2.

Codebook Design and Autoencoder-based Codeword Selection for RIS-assisted Communications

A one-bit beamforming codebook for Reconfigurable Intelligent Surface (RIS) is designed for indoor environments. Each codeword in a codebook is a phase pattern at RIS elements that reflects the signal to a desired angle. For a specific reflection angle, those codewords are computed that maximizes the signal-to-noise ratio (SNR) at the receiver. Since multiple codewords can maximize the SNR, an algorithm to reduce the size of the codebook is proposed.

Microexpression Recognition using High-Level Features PyTorch

Micro-expressions are the brief and involuntary facial expressions occurring within a fraction of a second. In this work, we proposed a Deep Learning-based network model that relies on high-level motion features for Micro-expression Recognition. (Code will be uploaded soon.)

Device Discovery Services - Avahi & Python

A RIS device discovery service for detecting RIS-specific controllers by Generic RIS controllers.

A Monte Carlo simulation for Wireless Channel Generation for RIS/IRS-assisted Communication in PyTorch

Generating training data for Deep Neural Network-based optimization problem for Channel Estimation problem for IRS-assisted communication systems.

Slow as Hell

Paranoia

Led Zeppelin - Stairway to Heaven (Cover)

Ruining the legendry song with the best of my abilities.

Live Shredding @ IIITNR

This is an excerpt from a much longer version.

Gilmorish