We examined the influence of context on the articulatory positions of the tongue, lips, and mandible using optical and marker-tracking systems and speculated articulatory coordination during the production of Japanese VCV sequence from the standpoint of current theories of speech production. The primary articulator for the consonant was comparatively invariant in the context. The secondary articulator was variant in the context. Influence of the preceding vowel on the secondary articulators for the con-sonant decreased monotonically and the influence of the following vowel increased in the same way. Thus, movements during the Japanese VCV sequence seem to consist of basic vowel-to-vowel gestures and comparatively invariant movement for the relevant consonant articulator.

View full abstract

The measurement of energy transported by bending vibration in a structure may provide useful information on the problems of vibration transmission in the structure and sound radiation from it. Measurements were performed for one-dimensional propagation of bending vibration in beams using two and four pick-ups. The errors caused by the use of finite difference approximations, differences in gain and phase among channels, and disparity in pick-up intervals were also taken into consideration. Since there have been few studies on near fields, we also investigated how far the near field extends.

View full abstract

Tensile stress is applied to the specimens of polymethyl methacrylate (PMMA) andpolystyrene (PS), using a mechanical device composed of screws and a lever, and its effect to the ultrasonic velocity is observed by water immersion ultrasonic sing-aroundmethod for both compressional wave and shear wave. The incidence angle is set to be zero for the compressional wave, or to be the angle appropriately above the critical angle of the compressional wave for the shear wave. Relative ultrasonic velocity varia-tion (RUVV) due to applied stress is linear for all cases, and the effect of RUVV for theshear wave is larger than that for the compressional wave. Next, mechanical anisotropy of injected polypropylene (PP) plate and burned extruded porous machinable ceramic plate are investigated by the similar inclined water immersion method. The relative ultrasonic velocity difference (RUVD) of the injected polypropylene plate ranges up to about 9%.

View full abstract

When a noise arises from several individual noise events, then its equivalent continuous A-weighted sound pressure level (L_AeqT) depends on the number of noise events (N), the A-weighted sound exposure levels (L_AE) and the equivalent continuous A-weighted sound pressure level of the background noise (L_AeqT). At least some of those quantities are random variables. Thus, the values of LAeca, measured within a constant time interval, are random as well. The results presented here make it possible to predict the mean and variance of L_AeqT in terms of some characteristics of N, L_AE, and L_AeqT.

View full abstract

It is the aim of this paper to develop a strategical method for the determination of SEA total loss factors (TLFs) and some of the coupling loss factors (CLFs) for built-up structures based on the recent development of in situ measurement techniques and the Statistical Energy Analysis theory. An obvious advantage of the method is that the determined TLFs would include the dissipation of coupling, and ‘automatically’ share the dissipation loss of the coupling to each coupled subsystem. The present strategy provides an applicable means for determination of total loss and some coupling loss factors of built-up structures for high frequency structure borne vibration and sound prediction. Satisfactory results of numerical experiments of the present strategy have been achieved.

View full abstract

The quick and simple training of learning vector quantization (LVQ) can produce classi ficationpower at least as high as that by a powerful, but complex classifier based on artificial neural networks. However, LVQ is a discriminative training algorithm for a distance classifier handling static (fixed-dimensional) patterns. Thus, an innovative process is required to apply this algorithm to dynamic (variable-durational) speech patterns. To meet this requirement, an HHM/LVQ hybrid algorithm was proposed which integrated HMM (Viterbi) segmentation with LVQ classification. However, this algorithm, using all the possible HMM models for segmentation, produces an enormous number of training tokens, making it difficult to apply to large-scale continuous speech recognition tasks. In this light, we present a new minimum-distortion segmentation (MDS)/discriminative classification hybrid algorithm. The MDS algorithm produces one segmentation and this is used in place of the many HMM segmentations. To make a proper comparison between the two methods we used as our discriminative classifier the same LVQ formulation. For clarity, we refer to this proposed algorithm as an MDS/LVQ hybrid algorithm. Results on the E-set task show that MDS/LVQ, with its significantly reduced training, can achieve discriminative power at least as high as HMM/LVQ.

View full abstract