Title:

Chopped: Audio Driven Video Segmentation

High school baseball coaches often spend significant time manually reviewing and editing game footage to analyze player performance, making the process inefficient, especially for full-length recordings. This project presents Chopped, a web-based application designed to automate the segmentation of game film into meaningful highlight clips. The system leverages multimedia data processing by combining speech-to-text transcription with structured event data to identify key moments within a game. During recording, users verbally mark events such as “strikeout” or “double,” which are later transcribed and mapped to timestamps in the video. These timestamps are used to automatically segment footage into clips corresponding to each detected event. The system was evaluated using full-length (~2-3 hour) baseball recordings with live announcer audio, as well as simulated event data to validate timestamp accuracy. Results show that the system reliably generates accurate clips when keywords are clearly spoken, with minimal sensitivity to background noise. While performance depends on consistent input, Chopped significantly reduces the need for manual editing and demonstrates a practical approach to automated sports video analysis.

Name

Comment*

Comments are viewable only by submitter

Captcha*