Files in this item



application/pdfECE499-Sp2019-chen-Brian.pdf (1MB)Restricted to U of Illinois
(no description provided)PDF


Title:Detailed study of speech recognition
Author(s):Chen, Brian
Contributor(s):Levinson, Stephen
Subject(s):Speech recognition
speech recognition classification
Abstract:This research project provides a very detailed comprehensive overview and experimentation of the speech recognition process. Generally, speech recognition can be broken down into three phases. The first phase is the sample and denoise stage (Endpoint Detection Technique), which helps us collect the signals and separate background noise from the actual information. The second phase is the feature extraction stage (Spectrogram, Filter bank, MFCCs), which help us convert the received time domain signals to meaningful, useful frequency domain information prior feeding into classification model. Between the second and third phases, there is an optional phase that people often do known as the data compression phase, which will also be discussed in this paper. Lastly the final stage is the classification stage (KNN, CNN), which classified a specific input signal to one of the possible classes. On top of the different phases, we will also look into the classification results in great detail and see whether factors such as minimum squared error between different classes or length of the signal can play a role in the classified result.
Issue Date:2019-05
Date Available in IDEALS:2019-06-13

This item appears in the following Collection(s)

Item Statistics