As a result of determining a macroscopic structure of tone quality by sensory evaluations and multidimensional scaling method, it seemed that the psychological space applicable to tonal descriptive terms is composed of three main attributes, namely LOUDNESS(loud-sorf), PITCH(high pitched-low pitched) and PLEASANTNESS(pleasant-unpleasant) mutually almost orthogonal. Moreover, four sub-attributes intersecting obliquely to the three main-attributes were found. These sub-attributes included CONSONANCE (clear-turbid), BRIGHTNESS(bright-somber), RICHNESS (rich-thin) and SMOOTHNESS (smooth-rough). As a matter of fact, these four sub-attributes depend on the three main-attributes(Fig. 10). For example, the "clear" quality of consonance is "soft", "high pitched" and "pleasant" sound, while the "turbid" quality is "loud", "low pitched" and "unpleasant" sound, and this result thoroughly corresponds to the CONSONANCE THEORY. The tonal sources used for the present hearing experiments consist of four groups. They are (1) 57 synthetic tones the physical values of which are distinctly grasped, (2) 97 musical instrument solo tones, (3) 44 tones by varying a part of the sound of "Scheherazade", which is the orchestral music with dominant string sections, and (4) 33 tones by varying a part of "Pictures at an exhibition", which is the orchestral music with dominant percussion and brass sections. As for the variation of above mentioned two orchestral sources, the quantitative variations such as spectrum form, intensity, phase modulation and echo were given, 41-53 hearing panels evaluated these tones by the choise of 38 tonal descriptive terms(Table 1) in the auditorium. For analysis, the program by Kruskal was used. (Fig. 4). As the result of this analysis, configurations in Fig. 6〜9 were obtained in three dimensions for each two source groups(Table 3). As for the distribution of the terms for these four source groups, the words of praise and displeasure distributed in two semispherical shell space divided by LOUDNESS-PITCH plane. At the center of the two hemispheric configurations, "pleasant" and "unpleasant" distributed, which represent important factors. This axis was stable even if the sources are different, therefore, PLEASANTNESS is regarded as the third main attribute following LOUDNESS and PITCH(Table 4). Next, the stable terms distributing in the diagonal quadrants among eight quadrants in Fig. 6〜9 are closely examined in common in four source groups, then the terms of four sub-attributes described above were extracted. When the above experimental data were reanalyzed with 14 terms concerning three main attributes and four sub-attributes, the stress in three dimensions decreased further, and approximately similar configuration was obtained. Based on these facts, it is suggested that the fourteen tonal descriptive terms concerning three main-attributes and four sub-attributes are sufficient for deriving the macroscopic structure of tonal sources.
In the previous report, it was suggested that the psychological space applicable to tonal descriptive terms is composed of three main attributes (LOUDNESS, PITCH, PLEASANTNESS) and four sub-attributes (CONSONANCE, SMOOTHNESS, BRIGHTNESS, RICHNESS)(Fig. 1). This paper reports on the results of studies on the characteristics of these seven attributes. Tonal source groups(pure tones, synthetic tones, musical instrument solo tones, and reproduced musical sounds;total 7 groups, 426 tones) were evaluated, respectively, by the absolute judgment using 38 tonal descriptive terms(Fig. 2). By multidimensional scaling method and other evaluaion results, the relations between three dimensional configurations of tonal sources(Table 1, Fig. 5, 6, 9, 10, 15, 17, 22) and tonal descriptive terms were analyzed for each group. In case of pure tones, in the project of configurations on PITCH-PLEASANTNESS plane, "arch" shaped curves were shown for each loudness(Fig. 5, 6, 9, 10), and the most "pleasant" parts of these curves are all nearly 1kHz. This configuration is called "TONE ARCH"(Fig. 5, 6). Moreover, the neutral points of PITCH (high pitch-low pitch) and LOUDNESS (loud-soft) depending on absolute judgment were found. The neutral point for the PITCH of pure tones and synthetic tones is about 600Hz, 700Mel(Fig. 3), and that for LOUDNESS is 70 phon (pure tones, synthetic tones, musical instrument solo tones) or 80 phon (reproduced music)(Fig. 7, 8). CONSONANCE(clear-turbid) and SMOOTHNESS(smooth-rough) are the attributed associated with the frequency intervals of tone components and its loudness, and in auditory nerves, CONSONANCE is concerned with "Space pattern", while SMOOTHNESS is concerned with "Time pattern". The boundary situation between CONSONANCE and SMOOTHNESS corresponds thoroughly to Beatty's "most rough frequency interval", and it is the problem different from critical band (Fig. 14, 15, 16, 17). BRIGHTNESS (bright-somber), similarly to the sense of sight, is related with the sensitivity to distinctness in tone quality. Well sensible sound (the sound with high pitch, loudness and other discerning ability) is "Bright tone"(Fig. 19). For example, loud tone of nearly 1〜3kHz is "bright", however, low-pitched tone is "somber" (Fig. 18). RICHNESS(rich-thin) depends on the volume of tonal information of sound stimulation obtained by auditory nerves. Similarly to tone "VOLUME", loud and low-pitched tone is "rich", while soft and high-pitched tone is "thin". However, tone in "unstable dynamic domain" and its reverberated tone are also "rich"(Fig. 20, 21, 22). For normal persons's ear, these seven attributes for tone quality seem to have common sensibility in the domains where the attributes are felt intensivery.
This paper describes an improvement in frequency characteristics of interdigital surface wave transducers (ISWTs) effected by modifying the electrode structure of the ordinary ISWT. These modified transducers (Divided electrode-pairs ISWT) are composed of several minor ISWTs connected electrically in series and acoustically in tandem as shown in Fig. 1. These minor ISWTs are formed by dividing the electrode-pairs of the ordinary ISWT into several segments. The improvement of frequency characteristics by modifying the electrode structure are effected by changing the number of electrode-pairs of minor ISWT or by changing the length of coupling lines between the individual minor ISWTs. The equivalent circuit of the divided electrode-pairs ISWT composed of 3-minor ISWTs can be expressed as a tandem connection of equivalent circuit for the ordinary ISWTs and coupling lines as shown in Fig. 2. Therefore, characteristics of these ISWTs may be obtained by calculating the tandem connection of Y-matrices. Fig. 3 indicates the insertion loss of the delay lines which are composed of two indentical divided electrode-pairs ISWTs having the coupling line of one wave length. If the number of electrode-pairs of the middle minor ISWT is less than that of those at both ends, side lobes of frequency characteristics can be suppressed. Moreover since the motional impedance of this transducer becomes greater than that of the ordinary ISWT, the insertion loss is diminished at high frequencies where the motional impedance decreases. Fig. 4 shows an applicable range of the divided electrode structure to diminish the insertion loss. Fig. 5 shows the effect of variation in the coupling line-lengths upon the frequency characteristics of output open circuit voltage. It is shown that the narrow bandwidth can be obtained by using the coupling line-length of an integer multiple of the wave length. In experiments, these ISWTs are fabricated on the piezoelectric ceramics plates by photo-etching, the electrode dimensions being shown in the respective figures. Fig. 7 and 8 show the insertion loss of delay lines using 3-divided electrode ISWT as the transmitting ISWT. The side lobes of divided ISWT was suppressed more noticeably than that of the frequency characteristics of the undivided ISWT. The divided electrode structure becomes useless to diminish the insertion loss at low frequencies, but the output open circuit voltage increases regardless frequency range (Fig. 9). Fig. 10 shows the characteristics of narrow bandwidth obtained by variation of coupling line-lengths between the individual minor transducers without increasing the number of the electrode-pairs.
It is known that a sound articulation can be estimated in the presence of a steady noise. In most circumstances, however, the environmental noise consists of time-varying sources, and the speech interference of such a noise cannot easily be estimated. Several papers have appeared on this subject, but almost no study on the Japanese language can be found. In this paper the interference of time-varying noise with speech listening is described on the basis of experiments, in which the sound and syllable articulation, and sentence intelligibility are investigated by using a synthesized time-varying noise. Synthesized noise has a spectrum similar to that of road traffic noise (Fig. 2), and a change in sound level of noise is characterized in this experiment by the condition that the level fluctuations occur randomly (Fig. 1) and maintain the normal distribution as a whole (Fig. 4). Moreover a change in level occurs stepwise. Fig . 3 shows a block diagram of our experiments. 120 uttered lists for syllable articulation tests and 32 lists for sentence intelligibility tests were prepared prior to the experiments. The syllable articulation list consists of 100 syllables and the sentence intelligibility list consists of 50 easy questions respectively. Four female students with normal hearing acuity were the subjects in the articulation tests and the intelligibility tests respectively, and two of them participated in both experiments. Speech samples were presented to the subjects through the headphones at 62 dBSPL in the articulation test and 64 dBSPL in the intelligibility test. Fig. 5 and Fig. 6 show the relations between sound or syllable articulation and the duration of the step D, where noise is held constant, for some ⊿L's which denote the excess of long time rms level of speech over the median of sound level distribution of noise. Fig. 7 shows the relation between sentence intelligibility and D. From these figures and the results of analysis of variance, the effects of D on the articulation and the intelligibility are said to be not so great as ⊿L and σ, which are described below. Fig. 8 and Fig. 9 show the relations between sound or syllable articulation and the standard deviation of sound level distribution of noise (σ). As seen from the figures, the articulation score has a tendency to decrease with the increase of σ, especially for greater ⊿L. Fig. 10 shows the relation between sentence intelligibility and σ. The intelligibility also seems to decrease along with the increase in σ. The relation between syllable articulation and sentence intelligibility is shown in Fig. 11, and the relation between syllable articulation and sound articulation is drawn in Fig. 12. From these figures, we can say that the sentence intelligibility is more influenced by sound level fluctuation of noise than the syllable articulation, while the relation between syllable and sound articulation in the presence of time-varying noise is much the same as that for steady noise. We can say again from Fig. 11 that the score of the intelligibility corresponding to a certain score of the syllable articulation is smaller for the greater σ than for the smaller σ. In other words, the connected speech is easy to suffer losses from fluctuating noise. Since the articulation decreases with an increase in σ as seen from Figs. 8 and 9, it is worth while to investigate into the relation between those scores and the equivalent sound levels (L_<eq>) of time-varying noise. Fig. 13 denotes it. A simple relation is not found between the articulation and L_<eq> of noise. Fig. 15 shows the sound articulations estimated from the experimental articulation in the presence of steady noise (Fig. 14) and occurence probabilities of sound levels in time-varying noise. Fig. 16 and Fig. 17 show the results of syllable articulation and sentence intelligibility calculated in the same manner as in Fig. 15. It can be said from those figures that the estimated values give a
Aural reflex (AR) is considered to have a function to attenuate the level of sound incoming to the inner ear. It can be one of the factors of the noise susceptibility, that is, individual differences observed in hearing damage caused by high intensity noises. In spite of many works on AR, these is little evidence to show the relationship between AR and the noise susceptibility. The purpose of the present study is to obtain fundamental data on this relationship for the normal ears. The subjects used are fifty males and fifty females, ranged in age from sixteen to thirty, with normal hearing acuity. The acoustic impedance of the middle ear was measured with an electroacoustic impedance bridge (Madsen, Type ZO-72). The magnitude, threshold, and latency of AR ware also measured for pure tones and for white noise. A typical pattern of AR for contralateral stimulation is shown in Fig. 2. The magnitude of AR increases as the stimulus level is intensified. Large differences in AR magnitude were observed between subjects. Fig. 7 shows the mean value of reflex thresholds and their distribution for each of the stimuli. The wide ranges in the thresholds indicate the exsistence of individual differences. The average threshold for white noise was lower than those for pure tones. The threshold for ipsilateral stimulation of 1kHz tone was about 10dB lower than that for constralateral one. Table 3 shows the mean values of reflex latency and their ranges for each of the stimuli. Large individual differences were also observed the reflex latency. The reflex latency was measured at some different levels of stimuli for 1kHz tone and white noise. The results summarized in Table 4, 5 and Fig. 9 indicate that the reflex latency decreases with increasing stimulus level. Furthermore, an positive correlation was found between the latency and the reflex threshold, as shown in Fig. 10. From these results, it is conceivable that a person having a high reflex threshold might be susceptible to intensive noise, especially to a sound with short rise time and short duration such as an impulsive noise, because of his prolonged latency and small magnitude of AR. Therefore, it may be concluded that the threshold of AR can be used as an indicater showing individual susceptibility to implusive noise.
This paper presents the experimental results on the acoustic power of flow-generated noise and the head losses of silencers and methods of estimating them. The experimental apparatus is schematically shown in Fig. 1. The PWL (the over all power level, re. 10^<-12> watt) was calculated by Eq. (1). The geometries of the test silencers are given in Table 1. Figure 2 shows the relation of SPL to the frequency when flow velocity was changed. The sound of high frequency is in propotion to 6-7th power of the mean flow velocity. Figure 4 illustrates the dependence of the PWL of flow-generated noise on the geometric parameters of silencers. It can be expressed from these experimental results that the PWL depends primarily on mean Mach number. The PWL increases with increase of the expansion ratio of cross sectional open area and of the diameter of the tail pipe, but it decreases with increase of the length of the chamber. It is shown in Figs. 8 to 13 that the variation of the PWL depends on the pressure fluctuation producted by the separation of flow at the inlet of silencers. Table 3 gives the PWL of noise generated by flow through the silencers. It is shown clearly that the PWL depends on the geometrical parameters of the silencers and the mean flow velocity. Figures 14 and 15 show the band PWL in relation to overall PWL as the relative level. The relationship between Strouhal number and the frequency is in accordance with Eq. (2). It may be seen that the power spectra of flow-generated noise show the maximum level around Strouhal number of about 0. 2, and above this frequency, it decreases gradually. Table 5 gives the experimental results of the head losses in the silencers. It should be noticed that the head loss ⊿P depends primarily on the mean velocity of air flow. The ⊿P increases with increase of the length of the silencers, and it decrease with increase of expansion ratio of the cross sectional open area.