The audio signal separation has been extensively studied due to its wide variety of applications. Along with the advancement of processor calculation performance in the last a couple of decades, the research focus has also extended to realising effective and feasible signal separation in practical environments where the problem becomes more challenging. In addition to classical linear filtering techniques, manipulating the spectral amplitude of a signal by applying a postfilter such as the Wiener filter is known to be an effective approach for practical applications. However the postfilter calculation requires the power spectral densities (PSD) of the signals of interest to be estimated beforehand. This article overviews methods for estimating the PSD of signals using spatial characteristics of their sources. Several practical applications that utilise the estimated PSD are introduced with some experimental results demonstrating the potential of the approach for solving challenging problems in practical applications.
This paper provides a phenomenological quantitative function of the perceptual simultaneity range (PSR) in terms of the frequency separation and a possible explanation. The PSR for two pure tones was measured with the frequency separation between the tones from 0.07 to 4.09 octaves with the lower frequency fixed at 200 Hz. In this study, listeners judged the simultaneity of the tones using the perceptual synchrony–asynchrony cue while the possible use of the perceptual fusion–separation cue (``oneness'' or ``twoness'') was eliminated. Results show that the PSR plotted against the frequency separation can be fitted to two segmented linear regression lines, one decreasing for small frequency separation and the other increasing for large frequency separation. These regression lines intersect around the critical bandwidth. Results also show no effects of the frequency separation on the singular interval or points, such as tonal consonance, musical consonance, and harmonic relations. These results suggest that the perception of simultaneity is mainly determined by the peripheral representation of the tone distance. We propose a possible explanation for the behavior of the PSR for small frequency separation by considering the mechanics of basilar membrane motion. However, the explanation for the behavior for large frequency separation is still unclear.
In this paper, we demonstrate that instabilities of the cochlear transmission-line model depend on numerical solutions. The transmission-line model approximates the fluid motion and the mechanical vibration in the cochlea. The mechanical vibration is enhanced by active cochlear feedback gain. For a realistic cochlea, spatial distribution of the feedback gain is varied randomly. However, most modeling studies set the gain to be constant as in an ideal cochlea because the spatial variation of the gain affects divergence of the calculation. To discretize the cochlear model for computation, the finite difference method and mesh analysis have been proposed. The finite difference method has been commonly used to solve the model represented in mechanical form; mesh analysis has been used to solve the model represented as an electro-acoustical circuit. This paper develops a state-space model of the cochlea for each of the two methods. The state-space formulation is well suited for testing instabilities of the model. As the result, both models show similar responses and stabilities under constant feedback gain (the ideal cochlea). On the other hand, with a randomly varied gain factor (the realistic cochlea), the model discretized by the finite difference method demonstrates greater instability than the model discretized by the mesh analysis.
This paper introduces an implementation of a compulsory course, ``Digital Signal Processing I,'' in a blended learning style. This subject is delivered by two lecturers as weekly parallel classes using the same e-learning contents including online quizzes for weekly homework. This sophomore-level basic course is compulsory for all the students of the Department of Computer Science and Electrical Engineering, Kumamoto University. The course has been assessed over several years from the students' activities on an e-learning system and the final record of this course as well as a questionnaire-based survey. The averages of the total grades for both classes are almost the same and the distributions of the evaluation scores have a similar pattern. These tendencies suggest the usefulness of the teaching style for maintaining the equivalence of the subject provided in parallel classes by different lecturers.
It is well known that the support conditions of a specimen can significantly affect the results of the impedance tube measurement. In theory, the specimen should be mounted under the ideal slip condition without any circumferential air gaps. However, this cannot be fully realized due to the inaccuracy of material cutting, thus causing a constraint or gapped condition. Furthermore, the specimen is normally mounted lightly in contact with the bottom of the tube, or with a back air layer. In this paper, the effect of realistic support conditions for poroelastic materials is investigated by numerical simulation with the finite element method. Under each condition, the behaviors of calculated values of normal-incidence absorption coefficient and transmission loss are discussed by comparison with theoretical values for the infinite-area material. It is shown that several mechanical and acoustical parameters of poroelastic materials can be estimated from impedance tube measurement under the circumferential constraint and gapped conditions.