This diploma project examines the human quality perception of guitars. This provides the background for development of objective algorithms to distinguish between musical instruments on a quality basis.
Three listening tests have been developed, using two different methods (serial and block by block presentation of stimuli). They have been realized and evaluated. It was asked, to judge the acoustical overall quality of selected classical guitars, by listening to recorded tone sequences (scale and melody).
Due to the use of binaural recordings in the listening experiments, it was necessary to compensate for the characteristic of the used headphones, using digital filters. Three methods for measuring the frequency response of the headphone (pictures below: dummy head, miniature microphone in blocked ear canal, and artificial ear) have been evaluated and one was found to be suitable.
Parameters like the musician, the room, the played sequence and the repetition have been included into the experimental design. The pictures below show a guitarist during the recording of the samples in two acoustically different rooms:
The complexity of human quality perception can be seen from a multitude of interactions between the mentioned factors. It was concluded, that hierarchical plans, which would reduce the necessary effort in listening experiments, can only be applied very restricted.
Interestingly, there was no significant difference between the quality judgement of guitarists and non guitarists.
You can download the diploma thesis here (in German):
|Merchel, S., Untersuchungen zur subjektiven und objektiven Bewertung und Beurteilung von Musikinstrumenten anhand von Solomusikstücken, Diploma Thesis, TU Dresden, Dresden, Germany, 2005.|
For more details on parametric methods see (in German):
|Abschlußbericht Musicon Valley, TU Dresden and Institut für Musikinstrumentenbau Zwota|