AMATH 482 Project Reports

This repository contains reports on solutions to a variety of applied problems using MATLAB and Python

Tracking a Marble from Ultrasound Data

Given noisy ultrasound data, this problem looks to denoise the data, enabling tracking a marble lost in the stomach of a dog. The solution draws on the Fourier Transform and Gaussian Filters and was performed in MATLAB using MATLAB functions.

Time-Frequency Analysis on Music Clips

This problem looks to filter music signals in order to determine the notes being played and when they are played with the goal of recreating the music score for the various songs. The Gabor Transform is used in addition to the Fourier Transform and Gaussian Filters and is completed in MATLAB.

Principal Component Analysis of Video Data

After tracking the motion of a paint can through video files, this report performs Principal Component Analysis on the motion data to determine the energy of each dimension of motion. Singular Value Decomposition allows us to factor the data and perform PCA in MATLAB.

Music Classification

This supervised machine learning problem collects music data, trains classifiers to determine thresholds and tests the classifiers on new music data. This problem was completed with Singular Value Decomposition and Linear Discriminant Analysis in MATLAB.

Neural Networks

This project uses the MNIST Fashion data set and works in Python to implement and train a fully-connected neural network and convolutional neural network. The networks are tested and their accruacy is compared