Adaptive Grouped Paired Comparison in subjective sound quality evaluation

In the field of subjective sound quality evaluation, Paired Comparison (PC) experiment is widely used. However, the weakness that time consumption grows with square of the number of samples makes the evaluation of large quantity of samples unrealistic. On the basis of Grouped Paired Comparison (GPC) in which improper reference samples may be chosen, An Adaptive Grouped Paired Comparison (AGPC) design is proposed to greatly decrease time required and to avoid the reduction of the precision by adaptively adjusting the reference samples or seeds by using prior information from results of last grouped comparison. In this study, the mathematic theory and implementation method were given, including the statistical model and optimality criterion. The performance of AGPC method was compared against that of conventional PC method through the computer simulations.