My name is Stephen William and I specialize in Speech Recognition, Audio Classification, and Natural Language Processing. Welcome to my Github!
Audio Classification for recorded speech using audio feature extraction and augmentation feed into a Convolutional Neural Network. The model has an accuracy of 70% and uses Librosa for audio preprocessing and Tensorflow for the Machine Learning models.
Music genre classification using the GTZAN Dataset and extracting important features like mel-spectrogram and MFCC. Multiple Machine Learning techniques where attempted and the best result was a model with 4 2 dimensional Convolutional Neural Netowrk layers each with MaxPooling and Batch Normalization. The model achieved a 73% training accuracy.
Speech Recognition experiments using the latest Deep Learning methods and feature extraction and augmentation techniques. I will be using popular speech datasets including TensorFlow's Speech Commands and the LibriSpeech ASR corpus. I will be using PyTorch for most of the audio data preprocessing and transformations as well as the Deep Learning models.