Speech Recognition in Dysphonic Patients
Author Information
Author(s): Muhammad Ghulam, Mesallam Tamer A, Malki Khalid H, Farahat Mohamed, Alsulaiman Mansour, Bukhari Manal
Primary Institution: King Saud University
Hypothesis
The study aims to evaluate the accuracy of automatic speech recognition (ASR) systems in recognizing speech characteristics of dysphonic patients.
Conclusion
The current ASR technique is not reliable for recognizing the speech of dysphonic patients.
Supporting Evidence
- 100% recognition accuracy was achieved for Arabic digits spoken by normal speakers.
- Recognition accuracy for dysphonic patients varied between 56% and 84.5% depending on the type of voice disorder.
- No significant improvement in ASR performance was observed after treatment for dysphonic patients.
Takeaway
This study looked at how well a computer can understand people with voice problems when they say Arabic numbers, and it found that the computer struggles to understand them.
Methodology
The study analyzed speech samples from 62 dysphonic patients and 50 normal subjects using a conventional ASR system with MFCCs and HMM.
Limitations
The ASR system showed significant loss of accuracy for dysphonic patients, and there was little improvement in performance after treatment.
Participant Demographics
Participants included 62 dysphonic patients and 50 normal subjects, all native Arabs aged 18 to 50.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website