A Speech Command Control-Based Recognition System for Dysarthric Patients Based on Deep Learning Technology
Problem
Dysarthric patients struggle to use voice control on mobile devices, and existing ASR systems require large computation, raising implementation costs.
Why it matters
Voice control is a key device interface, and dysarthric patients risk being excluded from it.
What the AI did
A convolutional neural network using phonetic posteriorgram (PPG) speech features (CNN–PPG) recognizes speech commands, replacing a heavier ASR-based system.
Results
93.49% accuracy vs CNN–MFCC 65.67% and ASR-based 89.59%; only 54% of the ASR system's parameter count.
Impact
The CNN–PPG system beat the ASR baseline in accuracy while using roughly half the parameters.
Record
| Venue | Applied Sciences |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Yu-Yi Lin, Wei-Zhong Zheng, Wei Chu, Ji-Yan Han, Ying-Hsiu Hung, Guan‐Min Ho, Chia-Yuan Chang, Ying-Hui Lai |
| Citations | 35 (OpenAlex, at import time) |
| References | 82 |
| Model or method | CNN–PPG (convolutional neural network with phonetic posteriorgram features) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.3390/app11062477 |
| Open access | gold |
| Keywords | dysarthric speech, speech command recognition, CNN, phonetic posteriorgram, assistive technology |
Importance rating
real but narrower advance
Improves voice-control access for dysarthric patients and cuts computation costs; a real but narrow accessibility advance, no scale stated.