This paper presents a new system for spoken Japanese digits recognition by a neural network using vocal tract shapes. The vocal tract shape is a suitable parameter for synthesis or recognition. The vocal tract shapes are used for the neural network as input data. We first propose a simple method by which the vocal tract shape is directly estimated from speech waves. A three-layered neural network is used in our recogni tionsystem. The network learning algorithms utilized here are conjugate gradient (CG) algorithm and backpropagation (BP) algorithm. Finally, we show the recognition results to prove the effectiveness of our method, and we show that the CG algorithm has several advantages compared to the BP algorithm.

View full abstract

A three-dimensional finite element method is applied to investigate the acoustic trans missioncharacteristics of vocal tract. The accuracy of the finite element solutions is assessed by comparing with the exact solutions for an infinitely long rectangular duct. Using the FE models of vocal tract during pronunciation of vowels, we examine the sound field in vocal tract and the effects of multi-modal propagation due to an interac tionof the plane wave mode and the higher order modes in high frequency region. The results of calculation show that the modal interactions become significant at frequencies above 5 kHz during pronunciation of vowels/a/, /u/and/o/, and that some dips appear in the transfer responses of the system due to the multi-modal propagation. It is shown that the traditional equivalent circuit model has a validity only at the frequencies below 5 kHz.

View full abstract

The mechanism of sound generation on the piano can be divided into four stages:-(a) transmission of energy from a hammer to a set of strings, (b) propagation of energy in the strings, (c) transition of energy from the strings to the soundboard through the bridge, and (d) acoustic radiation of the energy from the soundboard. This process is represented by an electrical circuit model, each string being represented by a transmis sionline, and the quantitative relationship of the model are calculated by a computer. When a hammer strikes a string, a nearly half-sinusoidal pulse is generated, and this propagates along the transmission line, and then is reflected by the impedances of the bridge and the fixed end of the string. The propagation and reflection change the waveform of the initial pulse and reduce its amplitude. The driving velocity of the soundboard is produced by applying the sum of the forces of all the strings to the driving pointimpedance. This driving velocity produces a sound pressure, and its waveform is determined by the transmission characteristics of the soundboard. An artificial piano sound can be produced from a calculated waveform through a D-A converter. When the fundamental frequency of each string in a set is slightly detuned, they produce beats in each partial making an inharmonic sound. The lower partials in the produced sound change relatively slowly, while the higher partials change relatively rapidly. The amplitude of the initial part of the sound ('initial sound') decays rapidly, while that of the sustained part ('aftersound') decays slowly. This explains how a delicate timbre of the piano sound is produced.

View full abstract

As changes in timbre of reflected sounds due to the difference of material of the boundaries are experienced, the elastic vibration of the boundaries may affect the reflected field. Thus it is required to take into account the effects of the elastic vibration of the boundaries. In this paper, the reflected sound field from an elastic plate with infinite extent is analyzed theoretically. The effects of properties of the plate and the angle of incidence of sound wave on the reflection characteristic of an elastic plate are discussed.

View full abstract

Lexical and coarticulatory effects on phoneme perception are investigated by measuring reaction times for the phoneme/k/at various positions in Japanese words and nonwords presented to six subjects as naturally spoken or as synthesized by concatenating moras. Facilitatory lexical effects are found both in the naturally spoken and concatenated words at the final or penultimate mora regardless of word length. Facilitatory and accumulative coarticulatory effects are found in naturally spoken stimuli but not in concatenated stimuli.

View full abstract

Vibrational characteristics of a point-driven “double shell” (two concentric submerged cylindrical shells coupled by the entrained fluid) are investigated. Strong-and weak-coupling theories are proposed to model the interaction of how the inner and outer shells are coupled by the entrained fluid. Each theory is based on Flugge's infinite-shell equations, the Helmholtz wave equation, and boundary conditions at the fluid-struc tureinterfaces which depend on the coupling situation. Strong coupling forms a stand ingwavefield in the entrained fluid; weak coupling, a free wavefield. Experiments are carried out where generalized near-field acoustical holography (GENAH) is employed to provide the experimental vibration characteristics in wavenumber-frequency space of the finite double-shell. It is confirmed theoretically and experimentally that the outer shell of the double shell exhibits two separate dispersion curves: A higher-frequency curve bears in-phase vibrations with respect to the inner shell and suggests a forced response of the inner shell; the other lower-frequency curve, out-of-phase vibrations, and an induced response of the outer shell. Fluid-structure coupling condition de finitelyaffects the lower-frequency curve, but barely the higher-frequency one. Com parisonbetween theoretical and experimental results demonstrates the validity of the strong-coupling model.

View full abstract