• Journal of the Korean Mathematical Society
• /
• v.60 no.1
• /
• pp.115-141
• /
• 2023

We study a class of left-invariant pseudo-Riemannian Sasaki metrics on solvable Lie groups, which can be characterized by the property that the zero level set of the moment map relative to the action of some one-parameter subgroup {exp tX} is a normal nilpotent subgroup commuting with {exp tX}, and X is not lightlike. We characterize this geometry in terms of the Sasaki reduction and its pseudo-Kähler quotient under the action generated by the Reeb vector field. We classify pseudo-Riemannian Sasaki solvmanifolds of this type in dimension 5 and those of dimension 7 whose Kähler reduction in the above sense is abelian.

• The Journal of the Acoustical Society of Korea
• /
• v.16 no.5
• /
• pp.112-121
• /
• 1997

In this paper, we study a model parameter compensation method for noise-robust speech recognition. We study model parameter compensation on a sentence by sentence and no other informations are used. Parallel model combination(PMC), well known as a model parameter compensation algorithm, is implemented and used for a reference of performance comparision. We also propose a modified PMC method which tunes model parameter with an association factor that controls average variability of gaussian mixtures and variability of single gaussian mixture per state for more robust modeling. We obtain a re-estimation solution of environmental variables based on the expectation-maximization(EM) algorithm in the cepstral domain. To evaluate the performance of the model compensation methods, we perform experiments on speaker-independent isolated word recognition. Noise sources used are white gaussian and driving car noise. To get corrupted speech we added noise to clean speech at various signal-to-noise ratio(SNR). We use noise mean and variance modeled by 3 frame noise data. Experimental result of the VTS approach is superior to other methods. The scheme of the zero order VTS approach is similar to the modified PMC method in adapting mean vector only. But, the recognition rate of the Zero order VTS approach is higher than PMC and modified PMC method based on log-normal approximation.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.18 no.12
• /
• pp.1970-1982
• /
• 1993

This paper describes an efficient array algorithm for parallel computation of vector-radix two-dimensional (2-D) fast discrete cosine transform (VR-FCT), and its VLSI implementation. By mapping the 2-D VR-FCT onto a 2-D array of processing elements (PEs), the butterfly structure of the VR-FCT can be efficiently importanted with high concurrency and local communication geometry. The proposed array algorithm features architectural modularity, regularity and locality, so that it is very suitable for VLSI realization. Also, no transposition memory is required, which is invitable in the conventional row-column decomposition approach. It has the time complexity of O(N+Nnzp-log2N) for (N*N) 2-D DCT, where Nnzd is the number of non-zero digits in canonic-signed digit(CSD) code, By adopting the CSD arithmetic in circuit desine, the number of addition is reduced by about 30%, as compared to the 2`s complement arithmetic. The computational accuracy analysis for finite wordlength processing is presented. From simulation result, it is estimated that (8*8) 2-D DCT (with Nnzp=4) can be computed in about 0.88 sec at 50 MHz clock frequency, resulting in the throughput rate of about 72 Mega pixels per second.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.441-446
• /
• 2020

In this paper, we propose the α-Feature Map Scaling (α-FMS) method which extends the FMS method that was designed to enhance the discriminative power of feature maps of deep neural networks in Speaker Verification (SV) systems. The FMS derives a scale vector from a feature map and then adds or multiplies them to the features, or sequentially apply both operations. However, the FMS method not only uses an identical scale vector for both addition and multiplication, but also has a limitation that it can only add a value between zero and one in case of addition. In this study, to overcome these limitations, we propose α-FMS to add a trainable parameter α to the feature map element-wise, and then multiply a scale vector. We compare the performance of the two methods: the one where α is a scalar, and the other where it is a vector. Both α-FMS methods are applied after each residual block of the deep neural network. The proposed system using the α-FMS methods are trained using the RawNet2 and tested using the VoxCeleb1 evaluation set. The result demonstrates an equal error rate of 2.47 % and 2.31 % for the two α-FMS methods respectively.

• Journal of Power Electronics
• /
• v.5 no.2
• /
• pp.151-159
• /
• 2005

The errors generated from current measurement paths are inevitable, and they can be divided into two categories: offset error and scaling error. The current data including these errors cause periodic speed ripples which are one and two times the stator electrical frequency respectively. Since these undesirable ripples bring about harmful influences to motor driving systems, a compensation algorithm must be introduced to the control algorithm of the motor drive. In this paper, a new compensation algorithm is proposed. The signal of the integrator output of the d-axis current regulator is chosen and processed to compensate for the current measurement errors. Usually the d-axis current command is zero or constant to acquire the maximum torque or unity power factor in the ac drive system, and the output of the d-axis current regulator is nearly zero or constant as well. If the stator currents include the offset and scaling errors, the respective motor speed produces a ripple related to one and two times the stator electrical frequency, and the signal of the integrator output of the d-axis current regulator also produces the ripple as the motor speed does. The compensation of the current measurement errors is easily implemented to smooth the signal of the integrator output of the d-axis current regulator by subtracting the DC offset value or rescaling the gain of the hall sensor. Therefore, the proposed algorithm has several features: the robustness in the variation of the mechanical parameters, the application of the steady and transient state, the ease of implementation, and less computation time. The MATLAB simulation and experimental results are shown in order to verify the validity of the proposed current compensating algorithm.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.45 no.4
• /
• pp.107-114
• /
• 2008

In this paper, we constructed a speech emotion recognition system that classifies four emotions - neutral, happy, sad, and anger from speech based on male/female gender discrimination. At first, the proposed system distinguish between male and female from a queried speech, then the system performance can be improved by using separate optimized feature vectors for each gender for the emotion classification. As a emotion feature vector, this paper adopts ZCPA(Zero Crossings with Peak Amplitudes) which is well known for its noise-robustic characteristic from the speech recognition area and the features are optimized using SFS method. For a pattern classification of emotion, k-NN and SVM classifiers are compared experimentally. From the computer simulation results, the proposed system was proven to be highly efficient for speech emotion classification about 85.3% regarding four emotion states. This might promise the use the proposed system in various applications such as call-center, humanoid robots, ubiquitous, and etc.

• Journal of Broadcast Engineering
• /
• v.16 no.6
• /
• pp.1069-1072
• /
• 2011

This paper proposes the method which reduces complexity and improves coding efficiency by solving a problem of HEVC bi-prediction. In current HM 3.0, it is frequently occurred that L0 motion information and L1 motion information are identical in blocks which are bi-predicted. In this case, L1 motion vector is replaced by non-zero motion vector which belongs to first available neighbor block of current block. If they are still identical, prediction mode is replaced by uni-prediction. As an experimental result, in LD(Low-Delay) case, decoding time is reduced roughly 2%~5% and coding gain is roughly 0.3%~0.5% compared with the HM 3.0 anchor.

• Journal of Korea Multimedia Society
• /
• v.4 no.3
• /
• pp.230-237
• /
• 2001

Adaptive filters are widely used for acoustic echo canceler, adaptive equalizer and adaptive noise canceler. Coefficients of adaptive filters are updated by NLMS algorithm. However, Coefficients are misaligned by ambient noises when they are adapted by NLMS algorithm. In this Paper, a method determined the adaptation constant by low-pass filtered instantaneous gradient vector of LMS algorithm using orthognality principles of optimal filter is proposed. At initial states, instantaneous gradient vector, that is the cross-correlation of input signals and estimation error signals, has large value because input signals are remained in estimation error signals. When an adaptive filter is conversed, the cross-correlation will be close to zero. It isn's affected by ambient noises because ambient noises are uncorrelated with input signals. Determining adaptation constant with the cross-correlation, adaptive filters can be robust to ambient noises and the convergence rate doesn't slower As results of computer simulations, it is shown that the performance of proposed algorithm is betted than that of conventional algorithms.

• The KIPS Transactions:PartB
• /
• v.11B no.6
• /
• pp.661-666
• /
• 2004

In the wireless communication systems, transmission errors degrade the reconstructed image quality severely. Error concealment in video communication is becoming increasingly important because transmission errors can cause single or multiple loss of macroblocks in video delivery over unreliable channels such as wireless networks and internet. Among various techniques which can reduce the degradation of video quality, the error concealment techniques yield good performance without overheads and the modification of the encoder. In this paper, lost image blocks can be concealed with the OBMC(Overlapped Block Motion Compensation) after new motion vectors of the lost image blocks are allocated by median values using the adaptive selection with motion vectors of adjacent blocks. We know our algorithm is more effective in case of continuous GOB loss. The results show a significant improvement over the zero motion error concealment and other temporal concealment methods such as Motion Vector Rational Interpolation or Median＋OBMC by 3dB gain in PSNR.

• Journal of Electrical Engineering and Technology
• /
• v.2 no.4
• /
• pp.468-477
• /
• 2007

The DTC of voltage source inverter-fed PMSMs is based on hysteresis controllers of torque and flux. It has several advantages, namely, elimination of the mandatory rotor position sensor, less computation time, and rapid torque response. In addition, the stator resistance is the only parameter, which should be known, and no reference frame transformation is required. The DTC theory has achieved great success in the control of induction motors. However, for the control of PMSM drives proposed a few years ago, there are many basic theoretical problems that must be clarified. This paper describes an investigation into the effect of the zero voltage space vectors in the DTC system and points out that if using it rationally, not only can the DTC of the PMSM drive be driven successfully, but torque and flux ripples are reduced and overall performance of the system is improved. The implementation of DTC in PMSM drives is described and the switching tables specific for an interior PMSM are derived. The conventional eight voltage-vector switching table, which is namely used in the DTC of induction motors does not seem to regulate the torque and stator flux in a PMSM well when the motor operates at low speed. Modelling and simulation studies have both revealed that a six voltage-vector switching table is more appropriate for PMSM drives at low speed. In addition, the sources of difficulties, namely, the error in the detection of the initial rotor position, the variation of stator resistance, and the offsets in measurements are analysed and discussed.