自适应分组成对比较法：原理及种子的选择

Adaptive grouped paired comparison:theory and selection of seeds

声品质主观评价实验中,大样本量的成对比较实验由于评价时间随样本数量的平方增长而变得准以实现,为此本文提出了自适应分组成对比较法。该方法不仅可以有效减少评价工作量并使得实验易于实现,而且通过对种子样本的调整,避免了普通的分组成对比较法由于关联样本选取不当而减小实验结果准确度的情况。论文给出了自适应分组成对比较法的数学原理和实现方法,提出了选取最佳种子的经验公式,最终通过计算机仿真和主观评价实验,与传统成对比较法的结果进行比较,证明了自适应分组成对比较法的有效性。

In the subjective sound quality evaluation based on paired comparison method, time cost grows with square of the number of samples, which makes the evaluation of large quantity of samples difficult. An Adaptive Grouped Paired Comparison （AGPC） design was proposed to greatly decrease time requirement and difficulties of subjective comparison test, and to avoid the reduction of the precision when improper reference samples were chosen in common grouped paired comparison by adapting the seeds. The mathematic theory and implementation method were given, including the experiential formula. It was shown from computer simulations and the subjective sound quality evaluation that the results of the proposed AGPC method were in good agreement with the true merits of samples and those obtained from traditional paired comparison.