One of the major reasons for using language models in speech recognition is to reduce the search space. Context-free grammars or finite state grammars are suitable for this purpose. However, these models ignore the stochastic characteristics of a language. In this paper, three stochastic language models are investigated. These models are 1) a trigram model of Japanese syllables, 2) a stochastic shift/reduce model in LR parsing, and 3) a trigram model of context-free rewriting rules. These stochastic language models are incorporated into the syntax-directed HMM-based speech recognition system, and tested by phrase recognition experiments. The phrase recognition rate is improved from 88.2% to 93.2%.

View full abstract

Formation process of an ideal soliton is discussed by using valid computer modeling for simulation. A solito a is formed under the condition of balance between nonlinear distortion due to finite amplitude effect and deformation caused by velocity dispersion effect. A model used in this paper is propagation in a thin fiber of fused silica and the velocity dispersion is given by Pochhammer solution. The simulation is done by intro ducingalternately finite amplitude effect and dispersion effect, without linear absorption. Results obtained are as follows:(1) A hyperbolic pulse makes its waveform change until it satisfies stable condition of soliton which is expressed by the relation between the pulse widthand the peak pressure, given by solution of K-dV equation and (2) after the pulse achieved the stable condition, it is not changed the waveform anymore and can propa gatewith the same waveform without attenuation of peak pressure.

View full abstract

Particle velocity and acoustic impedance density of the ultrasonic field on the axis of a circular flat transducer are computed together with sound pressure. In nearfield range, interference pattern with transducer distance is investigated. The phase delay of acous ticimpedance density jumps always from -π/2 to π/2, where amplitude becomes zero. But, this regular jump is not the case for sound pressure. Next, the ultrasonic field off the axis of the circular flat transducer is investigated. It is revealed that the case on the axis is the extremity. of the general ultrasonic field around the axis. The mean value over a model of receiving coaxial circular fiat transducer is also computed changing the ratio of the radius (α) of the circular flat transducer to the wavelength (λ) of the ultra sonicwave. Computed amplitude of sound pressure, particle velocity and acoustic im pedancedensity are tabulated with the normalized distance (zλ/a²). The mean ampli tudeof Z-component of particle velocity is always less than 1.0 and seems to be an appropriate response for the ultrasonic system of a pair of circular flat transducers.

View full abstract

A planar-structure acoustic focusing lens proposed by the authors is applied successfully to the scanning acoustic microscope as a substitute for the conventional concave lens. The lens is composed of a number of concentric annular grooves of uniform depth formed on the radiation plane according to the theory of a Fresnel zone plate. Because of the simple configuration with no spherical surface, the planar lens operating at a very high frequency can be fabricated easily using the conventional photolithographic tech nique.Good focusing capability of the lenses fabricated for operation at 100MHz and 200MHz has been demonstrated through the experimental and theoretical examinations.Images of satisfactory quality have been obtained using the scanning acoustic microscope equipped with the planar lenses.

View full abstract

Loudness comparisons were performed by four subjects, under two experimental conditions: free field (anechoic room) and diffuse field (reverberation room). Each subject adjusted the headphone level of critical band noise bursts until they were equally loud as those from a reference loudspeaker (70dB SPL). Measurement scatter was smaller in the diffuse field than in the free field. To examine the reliability of loudness judg mentsat high frequencies, another method-hearing thresholds by Bekesy tracking-was employed. Each subject's threshold was measured with both loudspeakers and head phones.After compensation was made for the loudspeaker and room transfer functions, headphone frequency response was extrapolated from the results. This method led to high-frequency responses similar to those from loudness comparison. A loudness comparison experiment in which the subjects continuously wore a headphone was performed. However, the method of sound pressure loss measurement should be reconsidered.

View full abstract