Performance of Isolated Digit Speech Recognition in Crowded Environment.
Speech recognition is a process that recognizes what the speaker says. Its objective is to extract, characterize and recognize the information in the speech signal conveying what the speaker says. One of major problems in speech recognition domain is disturbance caused by background noise. This dist...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
2007
|
| Subjects: | |
| Online Access: | http://etd.uum.edu.my/123/1/Muhamad_Arif_Hashim.pdf http://etd.uum.edu.my/123/2/Muhamad_Arif_Hashim-1.pdf http://etd.uum.edu.my/123/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Speech recognition is a process that recognizes what the speaker says. Its objective is to extract, characterize and recognize the information in the speech signal conveying what the speaker says. One of major problems in speech recognition domain is disturbance caused by background noise. This disturbance can decrease the effectiveness and reliability of the system and its accuracy. This research objective is to measure the
performance of isolated digit speech recognition in crowded environment. VQSR prototype uses two kinds of distance measure: Euclidean distance and city block
distance. Noisy digit speech, which is constructed from TIDigit speech database and cafeteria noise from CLSU database, is used to train and test the prototype. The
prototype is also tested using real data that been recorded in a crowded and noisy cafeteria. Results of training and testing phases are recorded and compared between these two distance measures using a set of performance measurement analysis. This set includes Sensitivity, Specificity, Total Accuracy, False Acceptance Rate, False Rejection Rate and Half Total Error Rate analysis. Based on the performance measurement, a robust and reliable digit speech can be used by user that has high
possibility of success and low probability in making errors. Finally, the proposed model and guideline in evaluating the digit speech performance can be use in other speech domain. |
|---|