Alex Yin [260502051], zixuan.yin@mail.mcigll.ca Bonnie Hu [260556970], guanqing.hu@mail.mcgill.ca
Simple speech command is a widely requested feature in many trending artificial intelligent applications. This project presents the implementation and testing of a Convolutional Neural Network (CNN) solution to classify simple spoken commands with tuning results of various speech preprocessing techniques.
The training set, including labels, as well as a test set were downloaded from Kaggle's TensorFlow Speech Recognition Challenge page : https://www.kaggle.com/c/tensorflow-speech-recognition-challenge .
We uses 10/30 words in the data set and achieves 96% accuracy on test set.
You can see details in our report.