• The Journal of the Acoustical Society of Korea
• /
• v.41 no.3
• /
• pp.326-334
• /
• 2022

In this paper, we propose speech enhancement algorithm as a pre-processing for robust speech recognition in noisy environments. Auxiliary-function-based Independent Vector Analysis (AuxIVA) is performed with weighted covariance matrix using time-varying variances with scaling factor from target masks representing time-frequency contributions of target speech. The mask estimates can be obtained using Neural Network (NN) pre-trained for speech extraction or diffuseness using Coherence-to-Diffuse power Ratio (CDR) to find the direct sounds component of a target speech. In addition, outputs for omni-directional noise are closely chained by sharing the time-varying variances similarly to independent subspace analysis or IVA. The speech extraction method based on AuxIVA is also performed in Independent Low-Rank Matrix Analysis (ILRMA) framework by extending the Non-negative Matrix Factorization (NMF) for noise outputs to Non-negative Tensor Factorization (NTF) to maintain the inter-channel dependency in noise output channels. Experimental results on the CHiME-4 datasets demonstrate the effectiveness of the presented algorithms.

• Transactions of the Korean Society of Automotive Engineers
• /
• v.17 no.4
• /
• pp.1-9
• /
• 2009

A simplified model of seat-human body is presented to analyze vibrations of human body on a seat of vehicle. The theoretical model having seven degrees-of-freedom is composed of the inter-connected masses, springs and dampers. Until now, evaluation of ride comfort has been usually performed only through vehicle tests. This study aims to complement shortcomings of conventional vehicle tests in evaluation of ride comfort by using the theoretical model. The acceleration values of the human body are obtained from frequency response functions of the theoretical model. Thereafter, Ride and SEAT indexes are acquired by considering response characteristics of the human body for the 12 axes that are presented in BS 6841. A vehicle test is carried out to measure the acceleration values for the three parts of the human body such as upper body, hip and foot. Ride and SEAT indexes of the vehicle test are also obtained by considering the response characteristics of the human body, of which results are compared with the values from the theoretical model. It is found that the theoretical results are in good agreement with the experimental results.

• The Journal of the Acoustical Society of Korea
• /
• v.24 no.1
• /
• pp.1-10
• /
• 2005

This paper presents a wideband LPA (local polynomial approximation) beamforming algorithm that is appropriate for wideband moving sources. The Proposed wideband LPA algorithm adopts STMV (steered minimum variance) method that utilizes a steered covariance matrix obtained from multiple frequency components in one data snapshot, instead of multiple data snapshots in one frequency bin. The wideband LPA cost function is formed using STMV weight vector. The Proposed algorithm searches for the instantaneous DOA and angular velocity that maximize the wideband LPA cost function. resulting in a higher resolution performance than that of a DS LPA beamforming algorithm. Several simulations using artificial data and sea trial data are used to demonstrate the performance of the Proposed algorithm.

• Transactions of the Korean Society of Automotive Engineers
• /
• v.15 no.2
• /
• pp.50-57
• /
• 2007

Shocks are excited by impulsive forces and cause discomfort in vehicles. Current standards define means of evaluating shocks and predicting their discomfort, but the methods are based on research with a restricted range of shocks. This experimental study was designed to investigate the discomfort of seated subjects exposed to a wide range of vertical shocks. Shocks were produced from the responses of one degree-of-freedom models, with 16 natural frequencies (from 0.5 to 16 Hz) and four damping ratios (0.05 0.1, 0.2 and 0.4), to a hanning-windowed half-sine force inputs. Each type of shock was presented at five vibration dose values in the range $0.35\;ms^{-1.75}$ to $2.89\;ms^{-1.75}$. Fifteen subjects used magnitude estimation method to judge the discomfort of all shocks. The exponent in Stevens' power law, indicating the rate of growth in discomfort with shock magnitude, decreased with increasing fundamental frequency of the shocks. At all magnitudes, the equivalent comfort contours showed greatest sensitivity to shocks having fundamental frequencies in the range 4 to 12.5 Hz. At low magnitudes the variations in discomfort with the shock fundamental frequency were similar to the frequency weighting $W_b$ in BS 6841, but low frequency high magnitudes shocks produced greater discomfort than predicted by this weighting. At some frequencies, for the same unweighted vibration dose value, there were small but significant differences in discomfort caused by shocks having different damping ratios. The rate of increase in discomfort with increasing shock magnitude depends on the fundamental frequency of the shock. In consequence, the frequency-dependence of discomfort produced by vertical shocks depends on shock magnitude. For shocks of low and moderate discomfort, the current methods seem reasonable, but the response to higher magnitude shocks needs further investigation.