The use of helium-rich mixture as a breathing gas solves physiological problems in the deep-sea or saturation diving, but it gives rise to a problem of "helium speech ", that is, degradation of speech intelligibility. Needless to say, the effective speech communication must be provided for making sure the life support and efficient action of divers in such a situation. However, the present state of art for overcoming the said problem is unsatisfactory. It is due to the lack of comprehensive studies covering human factors as well as the physical properties in helium speech. Especially, perceptual effects of the helium distortions have been almost unknown. The purpose of this article is to clarify the perceptual nature of helium speech through the articulation tests of Japanese CV-syllables and to study their relation to the physical nature observed by the acoustic analysis of helium speech. Speech materials were recorded in three experiments of the simulated dives of the Japanese Man-in-the-Sea Project (Seatopia). Experimental conditions and factors related to the articulation tests are listed in Tables 1 and 2. Result of analysis of variance on the factors of those tests in listed in Table 4. Among the helium distortions, the most important factor for perception is thought to be a nonlinear upward shift in formant frequency. This upward shift is due sound velocity elevation and the nonlinearity of it is due to increase in density. This relation is approximated by Eq. (1). The representative transposition curves are shown in Fig. 6. The main results are summarized as follows:1. Articulation scores of CV-syllables (Table 3 and Fig. 1) show that speech communication is not performed freely around 5 atm. and becomes almost impossible at a pressure greater than 10 atm. 2. The 5% rise of articulation score on the forth day in the five-day saturation dive (Table 5) is statically significant, which is probably due to the speakers' adaptation to the unusual voice environment. 3. Table 6 shows the confusion matrix of helium vowels. Its structure (Fig. 4(a)) is uni-directional and entirely different from that of the distorted vowels in normal air (Fig. 4(b)-(d)). 4. Confusions among consonants are listed in Table 7. Plosives, fricatives and nasals are difficult to recognize in this order. Confusions from voiced consonants to vowels and /j/, and those from unvoiced to /h/ are remarkable. Consonants which have front place of articulation show more errors than others having the same manner of articulation. 5. Comparison of our results with those in the cases of English (Table 9 and also see Table 1 and Fig. 6) verifies that perception of consonants having front place of articulation is greately affected by the nonlinearity in the formant shift. 6. For designing a helium speech unscrambler, it will be necessary to take account of the nonlinearity of the formant shift as well as the upward linear shift of the formant frequency.
This is a study on the field measurement method of absorption characteristics of acoustic panels by the correlation method using M-sequence signal. Particularly, the principle of the measuring apparatus is detailedly described. In this method, the absorption characteristics, i. e. the complex reflection coefficient is obtained from the correlated wave forms measured in the presence and absence of acoustic panel. The measuring apparatus, by which the equivalent incident and reflected sound waves can be measured, is composed of two parts; one is a M-sequence signal generator (Fig. 1), that gives a basic M-sequence signal and a delay signal, and the delay time can be set manually and swept according to the measuring time (Fig. 3). The other is field measurement apparatus mainly composed of multiple and time average circuits using the two signals mentioned above (Fig. 2). By a long enough delay time range and a unit step delay time accuracy of 2. 5μs, not depending upon the frequency, it is possible to measure the equivalent sound wave even in a high-audio frequency. Error of measurement of the correlation method and effects of the oblique incident absorption characteristics of panel and the cubic incident wave corresponding to the measuring condition of this method, on the measured values of normal incident absorption characteristics are discussed (Fig. 4, Fig. 5). Propagation sound waves for the reverse-square law (Fig. 6), incident and reflected sound wave s (Fig. 7), and absorption characteristics of lawn, soil and absorbent fence by this apparatus are shown (Fig. 8, Fig. 9).
It is often desirable to represent characteristics of a piezoelectric resonator by an equivalent circuit which has the same impedance as the resonator at a frequency range of interest. This is especially required when one apply the highly developed theory of LC filters to the design of piezoelectric filters. A resonator with high electromechanical coupling can be represented by a clamped capacitance (C_D), shunted by a number of motional arms as shown in Fig. 1. Each motional arm represents a resonance and consists of an inductance and a capacitance in series. Although this circuit representation is exact, tedious calculations are necessary to evaluate its impedance, and also its form is inconvenient for use in filter design. For a resonator with low coupling, a simplified equivalent circuit consisting of a parallel capacitance (C_0) and only one motional arm permits a good approximation. In fact, this equivalent circuit has been widely used for quartz crystal resonators. The consists of the motional arm can be determined in the same manner as the case of a low coupling resonator. A problem arise how to determine the parallel capacitance, if one is going to adopt the simplified circuit for an approximation. This is because the parallel capacitance C_0 now consists not only of the clamped capacitance C_D but also of the effects from other neglected motional arms and becomes highly frequency dependent. Hence a capacitance measurement at any one frequency will yield an ambiguous result. For example the use of the capacitance value at a very low frequency as the parallel capacitance yields a poor approximation of resonator impedance. This paper shows that the simplified equivalent circuit is still a reasonable approximation for a high coupling resonator, provided that parallel capacitance is so chosen that the circuit yields the exact antiresonance frequency. Since the antiresonance frequency can be easily measured, all the constants of the proposed circuit can be experimentally determined even when an exact solution is not available. This approximation was compared with known exact solutions for both piezoelectrically stiffened and unstiffened modes of vibrations. The accuracy was evaluated in two different ways. At first, the difference of the susceptance as a function of frequency was calculated. Fig. 3 shows the results for: (a) stiffened mode; (b) unstiffened extensional mode; and (c) unstiffened radial mode of a circular plate. The coupling factor (k) is a parameter. Next, the error of the shift of the antiresonance frequency due to load capacitance was calculated. Fig. 4 shows the results for: (a) stiffened mode; and (b) unstiffened extensional mode. Both results show that the proposed simplified circuit is accurate enough for most practical applications. As an example, the frequency characteristics of a lattice filter as shown in Fig. 5 were calculated based on the exact solution and on the present approximation. The two results shown in Fig. 6 were almost identical. As a comparison, a curve for the case that the low frequency capacitance is chosen as the parallel capacitance C_0 is included in the figure. It poorly approximates in contrast with the above results. A resonator in practice has various sources of loss. Its effects can be well approximated by the insertion of a series resistance in the motional arm, as it has been done for low coupling resonators. The authors with to thank Dr. S. Nishikawa for his careful reading of this manuscript.