• The Journal of the Korea institute of electronic communication sciences
• /
• v.9 no.5
• /
• pp.543-548
• /
• 2014

Multiple-model based speech recognition framework (MMSR) has been known to be very successful in speech recognition. Since it uses multiple hidden Markov modes (HMMs) that corresponds to various noise types and signal-to-noise ratio (SNR) values, the selected acoustic model can have a close match with the test noisy speech. However, since the number of HMM sets is limited in practical use, the acoustic mismatch still remains as a problem. In this study, we experimentally determined the optimal SNR mapping between the test noisy speech and the HMM set to mitigate the mismatch between them. Improved performance was obtained by employing the SNR mapping instead of using the estimated SNR from the test noisy speech. When we applied the proposed method to the MMSR, the experimental results on the Aurora 2 database show that the relative word error rate reduction of 6.3% and 9.4% was achieved compared to a conventional MMSR and multi-condition training (MTR), respectively.

• The KIPS Transactions:PartB
• /
• v.11B no.2
• /
• pp.207-212
• /
• 2004

This paper presents a reinforcement learning framework for stock trading systems. Trading system parameters are optimized by Q-learning algorithm and neural networks are adopted for value approximation. In this framework, cooperative multiple agents are used to efficiently integrate global trend prediction and local trading strategy for obtaining better trading performance. Agents Communicate With Others Sharing training episodes and learned policies, while keeping the overall scheme of conventional Q-learning. Experimental results on KOSPI 200 show that a trading system based on the proposed framework outperforms the market average and makes appreciable profits. Furthermore, in view of risk management, the system is superior to a system trained by supervised learning.

• Journal of the Institute of Convergence Signal Processing
• /
• v.11 no.4
• /
• pp.320-325
• /
• 2010

This paper is about the method and procedure to register the candidate sequentially in the face recognition system using the PCA(Principal Components Analysis). We use the method to update the principal components sequentially with the SKL algorithm which is improved R-SVD algorithm. This algorithm enable us to solve the re-training problem of the increase the candidates number sequentially in the face recognition using the PCA. Also this algorithm can use in robust tracking system with the bright change based to the principal components. This paper proposes the procedure in the face recognition system which sequentially updates the principal components using the SKL algorithm. Then we compared the face recognition performance with the batch procedure for calculating the principal components using the standard KL algorithm and confirms the effects of the forgetting factor in the SKL algorithm experimentally.

• Journal of Power Electronics
• /
• v.17 no.4
• /
• pp.892-904
• /
• 2017

A design of a Class-E inverter with only one inductor and one capacitor is presented. It is operated at the optimal operation mode for domestic cooker. The design principle is based on the zero-voltage derivative switching (ZVDS) of the Class-E inverter with a parallel load network, which is a parallel resonant equivalent circuit. An induction load characterization is obtained from a large-signal excitation test bench, which is the key to an accurate design of the induction cooker system. Consequently, the proposed scheme provides a more systematic, simple, accurate, and feasible solution than the conventional quasi-resonant inverter analysis based on series load network methodology. The derivative of the switch voltage is zero at the turn-on transition, and its absolute value is relatively small at the turn-off transition. Switching losses and noise are reduced. The parameters of the ZVDS Class-E inverter for the domestic induction cooker must be selected properly, and details of the design of the components of this Class-E inverter need to be addressed. A 1,200 W prototype is designed and evaluated to verify the validation of the proposed topology.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.53 no.3
• /
• pp.143-150
• /
• 2016

In this paper, we propose Korean monophthong recognition method optimizing muscle mixing based on facial surface EMG signals. We observed that EMG signal patterns and muscle activity may vary according to Korean monophthong pronunciation. We use RMS, VAR, MMAV1, MMAV2 which were shown high recognition accuracy in previous study and Cepstral Coefficients as feature extraction algorithm. And we classify Korean monophthong by QDA(Quadratic Discriminant Analysis) and HMM(Hidden Markov Model). Muscle mixing optimized using input data in training phase, optimized result is applied in recognition phase. Then New data are input, finally Korean monophthong are recognized. Experimental results show that the average recognition accuracy is 85.7% in QDA, 75.1% in HMM.

• Journal of the Institute of Convergence Signal Processing
• /
• v.19 no.4
• /
• pp.174-179
• /
• 2018

Currently, various security technologies such as fingerprint recognition and face recognition are being developed. However, although many technologies have been developed, the field of incorporating technologies is quite limited. In particular, it is easy to adapt modern security technologies into existing digital systems, but it is difficult to introduce new digital technologies in systems using analog systems. However, if the system can be widely used, it is worth replacing the analog system with the digital system. Therefore, the selected topic is the electronic attendance system. In this paper, a camera is installed to a door to perform a Haar-like feature training for face detecting and real-time face recognition with a Eigenface in principal component analysis(PCA) based face recognition using raspberry pi. The collected data was transmitted to the smartphone using wireless communication, and the application for the viewer who can receive and manage the information on the smartphone was completed.

• KEPCO Journal on Electric Power and Energy
• /
• v.4 no.2
• /
• pp.101-105
• /
• 2018

In this paper, the kernel regression model is applied for the case study of gas turbine abnormal state analysis. In addition to vibration analysis at the remote site, the kernel regression model technique can is useful for analyzing abnormal state of rotor vibration signals of gas turbine in power plant. In monitoring based on data-driven techniques correlated measurements, the fault free training data of shaft vibration obtained during normal operations of gas turbine are used to develop a empirical model based on auto-associative kernel regression. This data-driven model can be used to predict virtual measurements, which are compared with real-time data, generating residuals. Any faults in the system may cause statistically abnormal changes in these residuals and could be detected. As the result, the kernel regression model provides information that can distinguish anomalies such as sensor failure in a shaft vibration signal.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.10
• /
• pp.1290-1297
• /
• 2019

The paper proposes a distance estimation technique for ultra-wideband (UWB) systems using convolutional neural network (CNN). To estimate the distance from the transmitter and the receiver in the proposed method, 1 dimensional vector consisted of the magnitudes of the received samples is reshaped into a 2 dimensional matrix, and by using this matrix, the distance is estimated through the CNN regressor. The received signal for CNN training is generated by the UWB channel model in the IEEE 802.15.4a, and the CNN model is trained. Next, the received signal for CNN test is generated by filed experiments in indoor environments, and the distance estimation performance is verified. The proposed technique is also compared with the existing threshold based method. According to the results, the proposed CNN based technique is superior to the conventional method and specifically, the proposed method shows 0.6 m root mean square error (RMSE) at distance 10 m while the conventional technique shows much worse 1.6 m RMSE.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.4
• /
• pp.501-507
• /
• 2020

This paper proposes a new frame synchronization technique based on convolutional neural network (CNN). The conventional frame synchronizers usually find the matching instance through correlation between the received signal and the preamble. The proposed method converts the 1-dimensional correlator ouput into a 2-dimensional matrix. The 2-dimensional matrix is input to a convolutional neural network, and the convolutional neural network finds the frame arrival time. Specifically, in additive white gaussian noise (AWGN) environments, the received signals are generated with random arrival times and they are used for training data of the CNN. Through computer simulation, the false detection probabilities in various signal-to-noise ratios are investigated and compared between the proposed CNN-based technique and the conventional one. According to the results, the proposed technique shows 2dB better performance than the conventional method.

• Investigative Magnetic Resonance Imaging
• /
• v.25 no.1
• /
• pp.1-9
• /
• 2021

The aim of this study was to provide a new assessment of rotator cuff muscle activity. Eight male subjects (24.7 ± 3.2 years old,171.2 ± 9.8 cm tall, and weighing 63.8 ± 11.9 kg) performed the study exercises. The subjects performed 10 sets of the exercise while fixing the elbow at 90 degrees flexure and lying supine on a bed. One exercise set consisted of the subject performing external shoulder rotation 50 times using training equipment. Two imaging protocols were employed: (a) true fast imaging with steady precession (TrueFISP) at an acquisition time of 12 seconds and (b) multi-shot spin-echo echo-planar imaging (MSSE-EPI) at an acquisition time of 30 seconds for one echo. The main method of assessing rotator cuff muscle activity was functional T2 mapping using ultrafast imaging (fast-acquired muscle functional MRI [fast-mfMRI]). Fast-mfMRI enabled real-time imaging for the identification and evaluation of the degree of muscle activity induced by the exercise. Regions of interest were set at several places in the musculus subscapularis (sub), musculus supraspinatus (sup), musculus teres minor (ter), and deltoid muscle (del). We used the MR signal of the images and transverse relaxation time (T2) for comparison. Most of the TrueFISP signal was not changed by exercise and there was no significant difference from the resting values. Only the T2 in the musculus teres minor was increased after one set and the change were seen on the T2 images. Additionally, except for those after one and two sets, the changes in T2 were significant compared to those at rest (P < 0.01). We also demonstrated identify and visualize the extent to which muscles involved in muscle activity by exercise. In addition, we showed that muscle activity in a region such as a shoulder, which is susceptible to B0 inhomogeneity, could be easily detected using this technique.