This year project was part of the Embedded System class and especially the Internet of Things session. It focused on mastering elementary microcontroller programming, sensor integration and wireless MQTT protocol usage. The 4 students team I was part of decided to develop a greenhouse climate and light intensity control system, controlable by a remote device through connection to an MQTT broker.

Speech enhancement project

The objective of this Project was to implement an algorithm able to process audio data in real time, to suppress the background noise present in a speech recording. Background noise is unknown and can be of any nature, with any intensity. The program is written in C and runs on the TI DSK board provided.

The code does the operation described on the diagram and uses several filtering and transformation techniques to suppres background noise as much as possible while keeping the speech as clear as possible. In order to keep real time operations the program has to run as fast as possible.

Machine Learning Assignments

The machine learning module consisted of four challenging assignments. In each one we explored different techniques commonly used in machine learning: Perceptron Learning Algorithm for handwritten digit classification, Linear & Ridge Regression, Gradient Descent, Structural Risk Minimization, Features transformation, Cross Validation, Soft Vectors Machine, Principal Component Analysis.

White Board