1977 年 33 巻 3 号 p. 103-115
Many objective characteristics of loudspeakers are measured in factories and laboratories, but there is little knowledge about how they affect sound quality. It is necessary to reveal the relation between objective characteristics and sound quality in order to improve sound quality of loudspeakers. In this paper, the relation between subjective evaluation of sound quality and objective data to reveal the characteristics which influence sound quality to a great extent is investigated. In the subjective listening tests, preference judgments which were subjected to factor analysis, similarity judgments which were subjected to multidimensional scaling and verbal descriptions of sound quality were carried out. The factor analysis of preference data yielded a preference space of three factors (Fig. 5). The first of them is the "consensus preference" factor and the remaining two are the "individual difference" factors. As a result of multidimensional scaling, similarity data are summed up in three psychological dimensions (Fig. 7). Interpretation of these dimensions indicates that they represent "volume and extent", "brightness" and "beauty" respectively (Fig. 7). The relation between objective data and subjective data was analyzed in two respects. Firstly, measured data sixteen objective characteristics were rated from the view point of high fidelity reproduction (Table 3). These characteristics were fitted into the preference space as vectors (Fig. 14). Rating scores of sound pressure responses measured in a listening room and an anechoic room have a high correlation with the "consensus preference" factor. Secondly, similarity of response patterns among loudspeakers was calculated for each objective characteristic and it was subjected to multidimensional scaling. The configurations of loudspeakers based on objective similarity were compared with the ones based on subjective similarity (Fig. 16). Similarity of sound pressure response in the listening room has a close coincidence with subjective similarity. These results imply the necessity of measurements not only in an anechoic room but also in a listening room.