VoxAId
Detect. Understand. Improve
The problem VoxAId solves
VoxAId is an AI-powered speech screening tool designed to make dysarthria detection faster, easier, and safer for everyone involved.
For Speech Therapists & Doctors
- Quickly screen patients before detailed clinical evaluation
- Use AI-generated reports to support diagnosis
- Save time on repetitive initial assessments
- Track speech condition changes over time
For Caregivers & Families
- Check speech health at home using a simple voice recording
- Detect early warning signs before the condition worsens
- Share downloadable reports with doctors easily
- Reduce stress caused by uncertainty and delayed diagnosis
For Rural Health Workers & NGOs
- Screen patients in remote areas without specialist availability
- Provide low-cost, fast preliminary assessments
- Improve healthcare reach and early intervention
- Reduce dependency on expensive hospital visits
For Patients
- Safe, non-invasive early screening
- No medical equipment required
- Easy-to-use interface with one-click report download
- Helps decide when professional medical help is needed
How VoxAId Makes Existing Tasks Easier & Safer
- Replaces long, manual screening processes with AI analysis
- Eliminates the need for immediate specialist consultation
- Reduces late-stage diagnosis risks
- Improves decision-making with Explainable AI insights
VoxAId doesn’t replace medical professionals — it makes early speech care accessible to everyone.
Challenges we ran into
The model initially gave inconsistent predictions for real-world audio recorded on mobile phones.
Differences in sample rate, background noise, and long silences affected accuracy.
Some audio files caused inference errors due to unsupported formats or corruption.
We solved this by standardizing all audio to a fixed sample rate.
Added silence trimming and audio normalization before feature extraction.
Improved noise handling to focus only on speech-related features.
Tested the system using real phone-recorded voice samples.
After these fixes, the model became stable, reliable, and suitable for real-world use.
Tracks Applied (1)
Ethereum Track
ETHIndia
Technologies used