In this paper, we demonstrate how feedback from a saturation function can explain the generation mechanisms of cochlear distortion products (DPs) using a transmission line model of the cochlea, two simple nonlinear phenomenological models, and a power-law nonlinear model. The first model includes feedback via a saturating element that models outer hair cell motility toggled by mechanoelectric transduction. The second models focus on the saturation process in the cochlea, and result in either feed-forward or feedback. The third model can fit compressive growth in the cochlear input-output function. The models show compressive growth in input-output properties for a single tone and generate DPs at moderate input levels for two tones. The results of the transmission line model show a good fit to the experimental results. DP levels are concentrated when the levels of the two tones are equal in the simple feed-forward model and are widely distributed in the simple feedback model. The simple feedback model shows similar results to the transmission line model, with the exception of the dependence on the frequency rate. The results of the power-law nonlinear model with an appropriate power are comparable to the results of the simple feedback model. These results suggest that the saturating feedback process generates appropriate power-law nonlinearity and can account for the input level dependence of DP generation in the cochlea.
The net transport through the Seto Inland Sea, associated with the inflow of the Kuroshio Current from the Pacific Ocean, is the focus of this study. Direct measurement of the net transport was successfully accomplished during February–November 2012 using reciprocal sound transmissions in the Akinada Sea, the western part of the inland sea, while accurate transport was estimated only from mid-April to late July and from mid-September to October. The net westward transport averaged monthly for the accurately observed period of six months was 13,107± 2,544 m3s-1. This result implies that the inland sea water is exchanged with Pacific water every 778 days (2.1 years).
In this work, three experiments are conducted. In a realistic environment (freeway), a database for the noise dynamics versus velocity in a motorcycle helmet is computed. Next, under the controlled conditions of an anechoic chamber, the models for the feedback (FB) and feedforward (FF) paths in the helmet are obtained by a filtering and identification process. Finally, in the same chamber, a hybrid (FB+FF) robust controller is tested against the actual noise measured in the freeway, and it achieves attenuations between 20 and 30 dB.
This document presents an algorithm of switched orthogonalization of fixed-codebook (FCB) search in code-excited linear-predictive (CELP) speech coder and derivation of conditions for switching. Orthogonalization of FCB search is an early 1990's technology, and later efficient implementation was developed on some algebraic CELP (ACELP) speech coders standardized in 2000s. ITU-T Recommendation G.729.1 is such speech coder standardized in 2006. Orthogonalization of FCB search does not degrade CELP coder performance if ideal gain parameters are applied to codebook vectors. However, because of limited performance of gain quantization and overall optimization of CELP coding algorithm, the orthogonalization does not always give improved coder performance. This document presents a switched orthogonalization algorithm based on an estimated adaptive codebook (ACB) gain parameter, which is obtained through the orthogonalized FCB search. The algorithm was evaluated in G.729.1 coder. While the orthogonalization was switched off on 20% of voicing frames, segmental SNR was improved by around 0.1 dB in average.