Nikhil Kumar

This project aims to develop a machine-learning model capable of classifying five different cooking activities (brownies, pizza, sandwiches, salad, scrambled eggs) based on sensor data from an Inertial Measurement Unit (IMU) in the CMU-MMAC dataset. This dataset captures motion and orientation data from 38 subjects using a specific IMU sensor (ID 2796) across all subjects. The goal is to accurately predict the prepared recipe by analysing time-series data collected during the cooking process using deep learning technique called Recurrent Neural Networks (RNN).

Developed a wireless communication system
using STM32 microcontrollers to efficiently manage connections and data transmission with multiple devices.
Designed and implemented robust protocols for seamless wireless interactions. Optimized the system for reliable
performance in real-time applications.

Developed an Object Detection technique using Histogram of Oriented Gradients (HOG) in image processing applications like leveraging HOG features for robust detection of objects in images, with a focus on accuracy and efficiency. Demonstrated expertise in designing and implementing algorithms to detect objects within images through optimized sliding window Techniques accurately.

Software Used - Processing 4.3

This project is for paralyzed patients, Who depend on somebody even for basic needs. Here we introduce a wearable device, that will be placed on the patient's fingers according to the sign language the system will enable the voice announcement. For example, if he needs water he will bend the thumb, then the sound will be enabled like I need water”, similarly other fingers for other needs, and also a programmable physiotherapy device for paralyzed patients.

Developed a AI model for classifying handwritten
digits from the MNIST dataset using Convolutional Neural Networks (CNNs). This project involved data
preprocessing, model architecture design, training, and performance evaluation, demonstrating proficiency in
deep learning and image recognition techniques.

Implemented graph-based image segmentation,
partitioning images into meaningful regions using graph theory. Each pixel was treated as a node, with edges
representing intensity differences. The algorithm involved edge creation, sorting by weight, and segment merging
based on a threshold. This project enhanced image analysis, demonstrating accuracy, efficiency, and robustness
in various image types, contributing to advancements in computer vision.

The A3A8 algorithm is a cryptographic algorithm used in GSM (Global System for Mobile Communications) networks to authenticate and secure communication between mobile devices and the network.

 Software Used - MATLAB and SIMULINK.

Developed a
model to classify handwritten digits using the MNIST dataset, leveraging the Gaussian Naive Bayes approach.
The project involved preprocessing the dataset, training the model, and evaluating its performance. This work
demonstrated proficiency in implementing probabilistic classifiers and reinforced skills in image processing and
machine learning.

Implemented an SMS spam
detection system using character n-gram models. This project focused on analyzing text data to identify patterns
associated with spam messages. The approach included feature extraction, model training, and validation.
Through this project, I gained experience in natural language processing, specifically in text classification and
feature engineering.