Digital Signal Processing laboratory final project for detecting Bengali sign language from image frames using convolutional neural networks

This project detects sign language from image data using a convolutional neural network model running on MATLAB.

The project was ported from my own project Bangla-Number-Recognition-CNN which was originally implemented following a MATLAB example project titled "Speech Command Recognition Using MATLAB"

The sign recognition algorithm works as follows:

• Image frames are first acquired and preprocessed to create a dataset.
• In the training phase, the image dataset is fed to a CNN network along with the class labels, and the final loss function used is a binary cross entropy.
• New samples are passed through the network at runtime, and predictions generated with the trained network

Data collection

• Image samples was collected and saved to harddisk using the "Matlab code/data_acquire.m" script
• Image samples are converted to a .mat dataset using the "Matlab code/dataset_generator.m" script

Training Model

• "Matlab code/train_network.m" script was used to train the model and save the model parameters to another .mat file

Testing Model

• "Matlab code/test_network.m" script was used to load the trained network from the .mat file and make predictions from image stream of webcam