Intelligent Processing of Test Results
There are several problems related to the object evaluation process. First of all, experts should be trained and experienced in critical listening. An additional procedure based on a blind listening test provides an evaluation of the listener’s self-consistency. If a listener always ranks the same system in the same way, his reliability is thus undoubted. It is desirable for experts to be of the same background, i.e. acousticians, musicians, etc., but in practice it is not easy to find such a group that is willing to participate in a series of experiments. It should be remembered that listening sessions are quite tiring and time consuming. Moreover, it is not convenient to walk this group of experts from one acoustic interior to another, with the financial aspect also being of importance in this case. On the other hand, there is a practical way to carry out this experiment. The evaluation procedure can be based on simulations of hall characteristics. In this case, sound excerpts recorded in an anechoic chamber, thus without any reverberation, are used. These sound samples are then processed by adding, for example, some reverberation. An expert, while listening, is then asked to rate the performance using grades such as low, medium and high. A number of experts should take part in such evaluations, resulting in the relation of semantic descriptors to the particular parameter quantities. Furthermore, since the number of experts is rarely sufficient, this interrelation should also be validated statistically.
KeywordsMembership Function Listening Test Critical Band Reverberation Time Intelligent Processing
Unable to display preview. Download preview PDF.