Pitch-based musical performance scoring for a PC-based karaoke system

• Main Authors: Fabularum, Dann Ralston M., Margallo, Vergel Angelo C., Teng, Robertina Raina Karla B., Umlas, Robert Patrick A.
• Format: text
• Language: English
• Published: Animo Repository 2008
• Subjects: Karaoke; Music > Performance; Musical pitch; Voice frequency; Signal processing > Digital techniques; Computer systems; Digital electronics; Engineering
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/14386
• Institution: De La Salle University

Pitch-based musical performance scoring for a PC-based karaoke system is a study which integrates the general concepts in music with the theories and techniques in digital signal processing to produce justified criteria for judging a singer's musical accurateness and tonality through the use of time-frequency analysis tools. This thesis showcases an improvement to the standard karaoke machine, which uses loudness level of input voice signals as basis for performance scoring. The study includes the analysis of two inputs the MIDI file, which is a standard format used in most karaoke machines and digital music production and a test subject's singing voice. The analysis of MIDI files was aided through the use of a sequencer, which was used to extract the melody of certain MIDI files to be used as reference for the correct notes of the song to be imitated by the singer, and the MIDI toolbox for MATLAB, which is a compilation of specialized functions for the visualization of MIDI data. Analysis of the singing voice was accomplished by applying a short-time discrete Fourier transform for variable time frames of voice data which was based on the onset times and durations of notes from the reference melody, and by developing a technique in extracting the fundamental frequency component of each voice segment, which contains the pitch for the specific time frame. A scoring platform was developed using a certain criteria, which is geared toward fair judging, and a comparison between the detected notes of both the melody and voice signals was accomplished to produce the final output of the system, which is the musical performance score of the singer. The pitch detection algorithm was, through a series of tests involving comparison to standard references such as tuning forks and MIDI samples, deemed to be correct and accurate.