Fig. 9.10 Relationship between the calculated percentile of nonidentified syllables and that obtained by listening tests (r = 0.86,

p < 0.01)

A remarkable finding was that the most significant factor in the previous section was the effective duration of the ACF, τe. In this study also, the most effective and significant temporal factor was the τe value in the temporal factors. In order to obtain effects of the different direction of the noise disturbance, the spatial factors may be taken into consideration. Conclusions are as follows:

1.The syllable identification (NI) may be calculated by both temporal factors extracted from the ACF and spatial factors extracted from the IACF.

2.Particularly in the condition of this experiment, the value of τe as the temporal factor is the most significant as is similar to previous results (Ando et al., 1999); in addition, the WIACC in the spatial factor contributes significantly to the speech identification.

9.3 Effects of Sound Fields on Perceptual Dissimilarity

Perceptual dissimilarity is the perceived difference between the same sounds when they are either produced or heard at different locations. In this section, we discuss dissimilarity for the real sound field of an existing hall in relation to all of the temporal and spatial factors extracted from the ACF and the IACF, respectively. To incorporate dissimilarity into the model, two temporal factors ( t1 and Tsub) of the sound field are added to these temporal and spatial factors. At a given fixed seating position in a real room, the overall psychological distance between sound fields changing source locations on the stage can be obtained by keeping other sensory effects in a room constant. We shall show that the scale value of dissimilarity may be described by the linear combination of all the temporal and spatial factors of the sound field.

Previously, Yamaguchi (1972) carried out an experiment to empirically measure the perceptual dissimilarity between the same sounds heard from different seats in an existing concert hall. He demonstrated that two significant factors associated with listening position affected subjective similarity: the sound-pressure level and the reverberation characteristics. Edward (1974) also tested dissimilarity by studying differences between different halls and reported that the early-echo pattern, the reverberation time, and the volume level were the significant factors. Cocchi et al. (1990) and Sato et al. (1997) confirmed the effectiveness of the theory of subjective preference through investigations in existing concert halls. Sato et al. (2002) reconfirmed the effectiveness of the theory in an existing opera house as well.

In the current study, a reverberation-free, 4-s fragment of recorded orchestral music (Water Music, Suite No. 2 – Alla Hornpipe, composed by Handel) was used as a source signal (Hotehama et al., 2002). The music source was characterized in terms of the running ACF of the source signal after passing through an A-weighted network. An ACF analysis was carried out with an integration interval 2T = 2.0 s and running step-size of 100 ms, and factors (0), τe, τ1, and φ1 were extracted. As shown in Fig. 9.11, the minimum value of the effective duration of the source signal, (τe)min = 46 ms. This value is obtained at the most active piece of music and thus strongly related to preferred values of the temporal factors ( t1 and Tsub) of the sound field as discussed in Chapter 3 (Ando et al. 1989; Mouri et al., 2000).

Dissimilarity judgments were performed in a multiple-purpose hall, the 400-seat ORBIS Hall in Kobe, Japan, which is shown in Fig. 9.12. Six loudspeakers with identical characteristics were placed on the stage. Twenty student subjects participated in the experiment. They were divided into four groups and seated at specific positions in seating locations A, B, C, and D. To avoid the effects of other environmental conditions, each dissimilarity judgment was conducted at a fixed seat, and sound sources only were switched between the six source locations. Subjects judged difference as an overall impression between the paired stimuli. They were asked to rate the pairs on a subjective linear scale that had two extreme ends: “no different,” and “extremely different.” The judgment was made for 15 pairs of the six sound fields at each listener’s location. The interval between paired stimuli was 1 s. Each pair of sound fields was separated by a silent interval of 5 s, and the pairs were arranged in random order. Each session was repeated 5 times. To obtain the scale value of dissimilarity between sound fields, the original data of dissimilarity judgment were categorized into seven categories by the method of successive categories (Torgerson, 1958). The scale value of dissimilarity for each pair of source locations at the seat positions was obtained with all listeners as listed in Table 9.5.

To measure the acoustical factors extracted from the ACF and IACF, the music signal used in the dissimilarity judgments was reproduced from each loudspeaker. The signal was recorded at each listening position, through two microphones placed at the two ear entrances of a real person facing the center of the stage. To obtain binaural impulse responses, a maximum-length sequence (MLS) signal was reproduced from each loudspeaker (Alrutz, 1981).

Considering the fact that dissimilarity as well as subjective preference judgments may be made at the most “active and informative” running music piece, which indi-

Fig. 9.11 Factors of the running ACF of the source signal analyzed by 2T = 2.0 s with 100 ms of the running interval. (a) τe. (b) Relative(0), obtained as relative to the maximum value at

τ = 0.5 s. (c) τ1. (d) φ1

cates the minimum value of (τe)min extracted from the running ACF, values of τ1 and φ1 at the particular piece also were extracted. Values of LL, IACC, τIACC, and WIACC from the running IACF were also computed.

After obtaining the binaural impulse responses, values of t1 and Tsub were calculated. The value of t1 was defined by the time difference between the arrival time of the direct sound and that of the reflection, which is the maximum energy in the impulse responses. From the two measured values of t1 obtained at both ears, the one with the largest amplitude of the first reflection was selected as the t1 (Ando and Gottlob, 1979). The averaged value of Tsub of the 500-Hz and 1-kHz octave band center frequencies were applied here, because these frequency ranges are the dominant of the source signal. The measured temporal and spatial factors obtained by running the ACF, IACF, and binaural impulse response analysis are shown in Fig. 9.13. The factors extracted from the IACF were also chosen from

Fig. 9.12 Plan of the ORBIS Hall in which dissimilarity judgment was made. A–D: Locations of listeners. – : Source locations changed in the PCT conducted at fixed seating position during the judgment

Table 9.5 Scale values of dissimilarity judgments for each pair of source locations at seating positions of A, B, C, and D