Lê Chí Đại
← Projects
Machine learning · Speech · 2025

Spoken Digit Recognition

A team coursework project on recognizing spoken digits, with an audio feature pipeline, Hidden Markov Models, and an interactive demo.

RoleTeam member · ML4U coursework group
Year2025
ToolsPython · HMM · MFCC

Project visuals

Figures from project reports, pitch decks, and saved experiment outputs. Select an image to open it at full resolution.

Three speech waveforms paired with their MFCC feature heatmaps.
Waveforms and MFCC features for spoken-digit recordings. Source ↗
Gradio speech recognition interface showing an audio recording and a predicted digit.
The project’s Gradio spoken-digit recognition demo. Source ↗
A ten-class confusion matrix for spoken digits with reported accuracy of 93.78 percent.
Confusion matrix from the published coursework experiment. The figure reports 93.78% accuracy for that evaluation. Source ↗

The task

Recognize spoken digits from audio recordings. The project page documents audio exploration, feature extraction, HMM modeling, a Gradio demo, and evaluation.

Feature extraction

The published figures pair speech waveforms with MFCC representations, showing how audio is converted into feature sequences for the recognition model.

Team work & evaluation

Completed as part of the ML4U coursework group for CO3117 at HCMUT. The figures shown here come from the team’s published project page; they describe team results, not an individual contribution claim. The published confusion matrix reports 93.78% accuracy for its evaluation setup.

Next project

Talent Intelligence