ASR Course Work

This is a python implementation for speech recognition for Chinese numbers. It can recognize number from 0 to 9 in chinese. 这段代码实现了音频数字0~9的分类识别。

中文版

Environment

• python == 3.7
• numpy
• matplotlib
• sklearn
• librosa

Introduction

• First extract time-series characters from frames of time domain & frequency domain features.
• Then extract MFCC features using decimation-in-time radix-2 FFT algorithms
• Then use some basic machine learning algorithms (such as SVC, decision trees, etc.) to train a classification model
• Output accuracy and Confusion Matrix on test set
• Using ECOC methods to improve generalization ability