This paper reports real-time subjective listening tests to examine the relationship between the adaptive filter tap length of acoustic echo canceller and the acoustic environmental conditions. A real-time acoustic echo canceller simulation system, for the listening tests, includes two digital filter boards which calculate acoustic echo, pseudo echo, and output error signal by subtracting the pseudo echo from the acoustic echo. Subjects judge whether they could recognize the error signal or not by using several acoustic echo cancellers with tap lengths differ from each other. Reverberation times in the room, signal delay time in telephone line, and noise level in the received signals are varied as parameters. As a result: The tap length mainly depends on reverberation time and delay time in the telecommunication system. The noise affects mildly. And, the maximum range of temporal masking by subject’s voice is estimated as between 0.2 and 0.3 seconds.
Tone stopping is the act of ending the vibration of a piano string by the contact of the damper. This paper studies the perceptual effect of tone stopping. Five performances of a music passage were synthesized with piano tones of simulated tone stopping at sound level, i.e., with the tones that were obtained by processing the waveform of a single, sustained tone of a real piano to induce a desired ending profile with the onset portion kept intact. These performances were rated by ten musically trained subjects with the method of paired comparisons on twenty adjectives. The result indicated that: (1) a short plateau followed by a slow decay made the tone reverberating, lustrous, and beautiful, (2) a long plateau followed by a fast decay made the tone sticky, immature, and blunt, and (3) a short plateau followed by a fast decay made the tone tight, sharp, and nimble.
Vibration intensity is a useful quantity for indentifying vibration sources and propagation paths, which shows the magnitude and direction of vibration energy flow, and it has been calculated using multi-point vibration data through finite-difference (FD) method in previous studies. This report presents a method to estimate the structural intensity from the vibration at one point: the in-plane and out-of-plane vibration displacements. We apply this method to one-dimensional beam and two-dimensional plate, and compare the error with the conventional FD method. The proposed method has small error for estimating transmission power especially under weak standing wave condition in a beam. The authors also propose a simple device to display the magnitude and direction of transmitted power, which is based on the nature of vibration locus in flexural vibration. A small plate placed at the top of the device rotates in accordance with the transmitted power level.
In this study, we measure the vibrations of the unattached free violin top plate and back plate. We apply the pressure-based conformal holography with a hologram and a source surface coupling (PCHHS) as BEM-based acoustical holography in order to measure the vibration eigenmodes of the actual arching violin plates. We reconstruct and visualize the distributions of the sound pressure and the particle velocity on the surface of the violin plate from the measured hologram data in the sound field. The important vibration eigenmodes to know the acoustic characteristics of the violin plates are clearly reconstructed with good resolution.
Energy sources localization problem has been extensively elaborated in far-field scenario, but yet been fully addressed in near-field scenario. This paper presents an eigen-structure based near-field wideband sources localization method using arbitrarily spaced sensor array. First, we introduce a near-field sensor signal model, which takes into account both the attenuation of the sound pressure and the time delay of the signal. Far-field scenario could be considered as a special case of it. Estimation of the source locations is based on the straightforward exploitation of the eigenstructure of array power spectral density matrices. We use the analysis proposed by Wax et al. that locations are chosen as those whose location steering vectors are most nearly orthogonal to the set of eigenvectors belong to noise subspace over each frequency bin, but with a slightly different formation. Far-field sources localization could be considered as a 1-D (azimuth only) or 2-D (azimuth and elevation) problem. Our method solves the 3-D (azimuth, elevation and range) localization problem. A fast algorithm but with a sacrifice in the freedom in sensor arrangement is also presented. Simulation tests prove its validity and good performance.
In recent years, many applications such as speech recognition system and teleconference system are in the limelight. For these systems to perform well, it is necessary to get rid of the influence of surrounding noise. Many researchers have studied speech enhancement techniques. In the actual environment, the surrounding noise is not only time variant but also unmeasurable. Furthermore, the transfer function from source to sensing point is usually unknown. Therefore the noise canceller does not always effectively reduce the influence of the surrounding noise. In this paper, a new blind signal separation algorithm is proposed. This algorithm is designed for a 2-inputs 2-outputs system and the mixture matrix is represented using a trigonometric function. Computer simulations evaluate the performance of these algorithms.