A voice-activated word processor has been realized which provides text input by articulated, phrase-by-phrase Japanese speech, using an automatic learning algorithm to improve syllable recognition accuracy. A speaker-dependent, continuous-phrase speech recognizer incorporated into the word processor is capable of real-time recognition using 111 monosyllables as the basic recognition units. The recognizer is trained by new user to construct 590 reference syllable templates based on uttering a set of words, necessary for syllable template-matching process. Syllable templates are continually updated during use of the word processor through automatic learning. The automatic learning algorithm replaces low-accuracy syllable templates with new patterns extracted from the input speech. The replacement is sensitive to syllable context and is carried out based on recent and longer term history of the recognition accuracy of each syllable. Such learning information is derived from comparison between syllable recognition results and a user-confirmed character string of input phrase as correct, and is accumulated on a phrase-by-phrase basis. The automatic learning algorithm was tested in experiments using Japanese sentences read with pauses between phrases at approximately 4 to 5 syllables per second. The results, using eight speakers, show average syllable recognition accuracy of 82.5% with automatic learning, compared to 71.0% achieved without the learning. Further, the recognition accuracy is increased to 86.5% when the maximum number of syllable templates is increased to 2, 048; both template replacement and template addition are carried out until the maximum number of templates is reached.

View full abstract

The influence of a liquid sample layer on the non linearly generated second harmonic component is examined by computations and experiments. The sample with plane surfaces is inserted in the focal region normal to the axis, whose acoustical properties are different from those of surrounding medium (water). The theory is based on the quasilinear parabolic approximation, valid for the case of weak distortion and small concavity. A linear transmission through the boundaries is also applied for the second harmonics. The analysis for a source with a Gaussian profile ex citation shows that, in the case of a loss less sample, the second harmonic component generated at the sample can be the greater part of the second harmonic sound propagating in the post focal region. In the case of a lossy sample, a newly introduced quantity C_s evaluated from the waveform at the post focal region corresponds well to the sample properties excepting the linear absorption. The experimental results suggest the usefulness of the quantity C_s.

View full abstract

An L_eq is usually calculated through an averaging operation of anti-logarithm of the measured instantaneous level value in a decibel scale. Considering the nonlinear transformation of level observation, there seems to remain a principal problem that the average energy can not be exactly obtained owing to the inequality of quantization width in an energy scale differing from the equality of quantization width in a level scale. In this paper, the averaged sound energy is estimated on the basis of level statistics with help of the moment generating function of sound level by considering the fundamental relationship between the sound level and its energy. Then, an evaluation method of L_eq is derived in the series expansion form reflecting the stationary and Gaussian properties as the first expansion term and the correction term for the deviation from these statistical properties as the higher order expansion terms. Finally, the proposed estimation method is experimentally confirmed by applying it to the actual noise environment inside and outside a room.

View full abstract

An analysis is presented for the free vibration of shells with a variable meridional profile. The applicability of the shell theory is assumed and the governing equations of vibration of the shell are written as a matrix differential equation of the first order by using the transfer matrix of the shell. Once the transfer matrix has been determined by numerical integration of the matrix equation, the frequency equations of the shell are derived with terms of the matrix elements from the boundary conditions. The method is applied to shells of two types, and the results of numerical calculations of the natural frequencies and the mode shapes of vibration are presented.

View full abstract

In Japan many self-governments or research groups carry out surveys on noise and vibration. But mostly the raw data of the surveys are not opened to public and are not accumulated efficiently. It is necessary for researchers to compare data of some surveys. So we are assembling data of social surveys on noise and vibration, and constructing a database on computer system. Some data in the database are obtained by ourselves, others are presented by other research bodies. Files in the database are concerning with personal noise exposure, environmental noise, noise and vibration of railroad etc. In this report, each survey and contents of the files are explained. Also the process and problems of data storage and accumulation are discussed.

View full abstract

In this report, a statistical method for estimating generally the noise level probability distribution from the observed values of L_eq and specific L_x noise levels is proposed theoretically by introducing newly a statistical orthonormal expansion expression on an intensity scale taking the lognormal distribution as the first expansion term. The effectiveness of the proposed method is experimentally confirmed by applying it to the actual road traffic noise data through the comparisons of estimation accuracy between the proposed method and the previous one based on the statistical Laguerre expansion expression.

View full abstract