In this paper, an artificial neural net architecture for adaptive speech analysis and synthesis is proposed. Based on discussing the signal analysis procedure with mathematical method, the simulation experiment is performed firstly on a typical signal (a square wave). Compared with Fast Fourier Transform, some important characteristics of the Neural Net Signal Analyzer-Synthesizer (NNSAS) are drawn from the experiment results. Then, NNSAS is used for speech signal processing, which results in a new approach to speech analysis and synthesis. In terms of the auditory masking effect, only some major components of connection strength vector are used to produce synthetic speech. Under the high compressive rate conditions, NNSAS can still produce synthetic speech with high articulation.

抄録全体を表示

This paper is concerned with an experimental modal analysis of a loudspeaker system. Based on Frequency Response Functions (FRFs) measured in about 500 points uniformly spaced on the front panel of the loudspeaker system several vibrations modes of the loudsneaker membranes have been found. Then, the vibrations of the loudspeaker membranes were compared with vibrations of the cabinet of the loudspeaker system. It has been shown that the vibrations of the loudspeaker case were very small relative to the vibration of the loudspeaker membranes. Thus, the loudspeaker cabinet, which may be considered as a set of vibrating plates, is rather a source of a very weak sound and may be neglected. It does not influence significantly on the acoustics field radiated by the loudspeaker system.

抄録全体を表示

Hydrophone signals obtained by a Sea Beam 2000 system were analyzed using the Synthetic Aperture (SA) technique to obtain a high-resolution profile of seafloor topography. The survey line spans 5km with a water depth of 500 m to 1, 000m in Sagami Bay. Careful and slow navigation was performed to suppress deflection from the planned survey line. The data taken by 42 hydrophones (half of the hydrophones on the array of the Sea Beam 2000) were recorded for one hour at a 48kHz sampling rate. Simultaneously, to measure the deflection from the planned line, the roll, pitch, and heading angle of the ship and its positiond determined by Kinematic GPS were recorded. The reference orbit axis of the projector array was obtained first. Each hydrophone signal in the time domain was then corrected by time shifting at each designed beam angle in order to construct multiple received beam forms with no fluctuations on the reference orbit axis. This correction canceled the deviation of the projector array at transmission and the hydrophone when received on each received beam axis from the reference orbit axis. Finally, an SA beam was formed using three-ping and five-ping received beams with no fluctuation on the reference orbit axis. As a result, the resolution of the SA beams increased.

抄録全体を表示

In order to investigate the influence of the sound intensity probe itself on the accuracy of measurement, a diffraction of the plane progressive wave by two finite-length cascade cylinders is calculated by the boundary element method (BEM). The sound pressure distribution around the cylinders is compared with the measured one. Both results agree well with each other. Frequency characteristics of the force that acts on the surface facing the gap between two cylinders are calculated for various probe model geometries. The force response shows deviations at high frequencies where the probe size becomes comparable to the wave length. This deviation mainly depends on the diameter of the probe.

抄録全体を表示

A conducting recognition system has been manufactured experimentally that follows human conducting and controls musical performance in real-time. Since most conducting elements can be directed with only the right hand, the system recognizes important and universal right-hand conducting elements conforming to the conducting grammar: i. e., beginning and ending of a musical piece, beat timing including Agogik (tempo rubato), the beat number in a measure, fermata, dynamics, some aspects of articulation. Actual orchestras produce sound with a specific delay after the corresponding conducted beat. The system, which simulates this delay, can be profitable for unskilled conducting students. The conducting gesture is captured by low cost accelerometers and recognized by HMM (Hidden Markov Models). Since HMM can cope with both time structure variance and the gesture parameter characteristics variance, it produces effects in conducting recognition. Furthermore, musical score information and playing performance information are also applied to the recognition on the fuzzy model of musician's recognition processes. Since there are analogies between conducting and speech recognition, a comparison between them is also explained. The system conforms to the basic conducting grammar which has been systematized through historical polish and is used by orchestras all over the world.

抄録全体を表示