Norbert Eidenberger
Supervision: Univ.-Prof. Dipl.-Ing. Dr. Bernhard Zagar
Many companies still own audio data which is stored on analogue storage media. The quality of this audio data diminishes with increasing age of the analogue storage medium. Therefore there exists an interest in digitizing the audio data which will then be stored permanently. A problem is the evaluation of the quality of this digitized data. Following a request of a company a scientific project has been started. This project studies the various factors and errors which influence the quality of the audio data.
llowing a request of a company a scientific project has been started. This project studies the various factors and errors which influence the quality of the audio data. The objective of this project is the development of software that analyzes the digitized audio data regarding the errors. Based on this analysis the quality of the audio data is evaluated (see Figure 1). This evaluation result is then to be stored in a meta file which will be appended to the digital audio file.
This students project deals with a part of the problems specified by the client. The required problems are the calculation of the azimuth (angular misplacement of the playback unit of a playback device for magnetic tapes), the detection of the phase correlation (deviation from the stereo balance), the distinction between mono and stereo signals, the assessment of the loudness of a signal, the calculation of the audio level and hum detection.
Algorithms for these problems were developed using Matlab. Digital and statistical signal processing methods were used. This includes the cross power spectral density, the coherence function, resonators and digital weighting filters in the time and in the frequency domain. The algorithms were tested using digital data samples provided by the client. The data has also been used to improve the accuracy of the algorithms and to verify that the algorithms work correctly. For every problem a detection algorithm was found. The finished algorithms will be implemented in C/C++ in the course of the project. But this implementation is not part of this project.
Link: Audioinspector, opens an external URL in a new window
May 10, 2007