Analysis of digital audio samples for the extraction of musical information

TR Number

Date

1995

Journal Title

Journal ISSN

Volume Title

Publisher

Virginia Tech

Abstract

This project led to the development of a Macintosh application that analyzes digital audio samples and generates MIDI data. The program, compiled with THINK C++ 6.0, uses standard Macintosh 'snd_' resources as input. The output is written to a Standard MIDI File. The Macintosh application derived from this project behaves like any standard Macintosh application with windows, menus, and an "about box". The complete source code for the application is included in this report.

The Fast Fourier Transform (FFT) was considered for the transformation from the time domain to the frequency domain. This did not prove to be the most appropriate method for this application. A new method was developed that looks at the waveform and matches repetitions of the waveform to get the base frequency for the sound. The new method does not suffer from the problems involved in using the FFT. It can be used for the detection of a base frequency in a sample that has a waveform that repeats regularly and does not contain much noise. This is much faster to compute and yields a much more accurate frequency measurement as well as greater timing accuracy.

Description

Keywords

Citation