Speech Trainer

AI-Powered Presentation Skills Enhancement Platform

Back to Portfolio

The Problem

Public Speaking Anxiety

Public speaking is a fear that about 74% of people share. However, public speaking has a tremendous impact on our lives. People who are better speakers are generally more successful and have happier lives.

The challenge is: How can a computer analyze the quality of a speech automatically?

Our Solution

Speech Trainer is a machine learning-based software application that provides users with comprehensive feedback scores to improve their presentation skills.

By analyzing six key presentation qualities, we help users develop confidence and effectiveness in public speaking.

Solution Overview

I researched several TED talk videos and noticed that verbal delivery and body language are very important to make a presentation effective. From this research, I identified six key presentation qualities that significantly impact speech effectiveness:

Posture: Facial landmark tracking
Filler Words: "um", "like", "basically" detection
Talking Speed: Words per minute analysis
Clarity: Word pronunciation assessment
Pause Quality: Length and frequency analysis
Confidence: Overall presentation scoring

Implementation

1
Research & Analysis
Identified six presentation qualities and wrote Python functions to evaluate any user video on these metrics.
2
Data Collection
Collected 30-second clips from TED talks and other prominent speakers to train the machine learning model.
3
Model Development
Created a machine learning model that predicts presentation quality scores for any headshot video.
4
Web Application
Built a website that evaluates and scores user videos on the six presentation qualities in real-time.
Speech Trainer Project Architecture

System architecture showing the flow from video input to scoring output

Technologies Used

OpenCV: Facial landmark detection and computer vision
PostgreSQL: Data storage and management
Pandas: Data manipulation and analysis
Scikit-learn: Machine learning algorithms
Psycopg2: PostgreSQL connection library
Jupyter Notebook: Interactive data analysis
Seaborn: Data visualization and graphing
Science Fair Success
Participated in the Alameda County Science and Engineering Fair (ACSEF) and won Second Place for this innovative project.