Projects Guided
- System identification using Machine learning/ Deep Learning approach
Description: We need a model to analyze the process to do different operations. The conventional methods use the first principal approach to identify the model of the process. The first principal approach uses the laws of physics to identify the model but the processes are highly complex and nonlinear. So, the better approach is data driven approach. In this approach the input -output data is used for fitting the model.
The reported data driven approaches use generally parametric approaches which does not capture all the necessary characteristics of the process.
In this work, both the non-parametric and parametric approach is described because non- parametric approach gives a lot of insights about the process. The machine learning approach is used for estimate the model parameters.
​
- Character recognition using radial basis neural network
Description: Identifying almost identical characters is one of the issues that have arisen in the study of handwritten scanned images of documents. As a result, the domains of pattern recognition and image processing gave it attention. There are two steps in the pattern recognition system.
Feature extraction is the first step, while classification is the second, In order to identify handwritten English characters.
The Radial Basis Function (RBF) networks are employed for the classification in this work.
​
- Emotion recognition using speech.
Description: The goal in this project is to categorize the speech as belonging to one of the three emotions—fear, anger, or sadness.
The samples used to perform this project are taken from authentic sources. Energy, pitch, MFCC coefficients, LPCC coefficients, and speaker rate are the key features identified from the samples
The classification is employed by the Support Vector Machine (SVM) classifier to categories these emotional states.