• ETRI Journal
• /
• v.32 no.5
• /
• pp.795-800
• /
• 2010

Embedding a large vocabulary speech recognition system in mobile devices requires a reduced acoustic model obtained by eliminating redundant model parameters. In conventional optimization methods based on the minimum description length (MDL) criterion, a binary Gaussian tree is built at each state of a hidden Markov model by iteratively finding and merging similar mixture components. An optimal subset of the tree nodes is then selected to generate a downsized acoustic model. To obtain a better binary Gaussian tree by improving the process of finding the most similar Gaussian components, this paper proposes a new distance measure that exploits the difference in likelihood values for cases before and after two components are combined. The mixture weight of Gaussian components is also introduced in the component merging step. Experimental results show that the proposed method outperforms MDL-based optimization using either a Kullback-Leibler (KL) divergence or weighted KL divergence measure. The proposed method could also reduce the acoustic model size by 50% with less than a 1.5% increase in error rate compared to a baseline system.

• Journal of Institute of Control, Robotics and Systems
• /
• v.16 no.12
• /
• pp.1137-1142
• /
• 2010

This paper studied the speech parameters less affected by the human emotion for the development of the robust speech recognition system. For this purpose, the effect of emotion on the speech recognition system and robust speech parameters of speech recognition system were studied using speech database containing various emotions. In this study, mel-cepstral coefficient, delta-cepstral coefficient, RASTA mel-cepstral coefficient and frequency warped mel-cepstral coefficient were used as feature parameters. And CMS (Cepstral Mean Subtraction) method were used as a signal bias removal technique. Experimental results showed that the HMM based speaker independent word recognizer using vocal tract length normalized mel-cepstral coefficient, its derivatives and CMS as a signal bias removal showed the best performance of 0.78% word error rate. This corresponds to about a 50% word error reduction as compare to the performance of baseline system using mel-cepstral coefficient, its derivatives and CMS.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.5
• /
• pp.2414-2433
• /
• 2019

The aim of this paper is to design a 3D laser radar prototype based on laser triangulation. The mathematical model of distance sensitivity is deduced; a pixel-distance conversion formula is discussed and used to complete 3D scanning. The center position extraction algorithm of the spot is proposed, and the error of the linear laser, camera distortion and installation are corrected by using the proposed weighted average algorithm. Finally, the three-dimensional analytic computational algorithm is given to transform the measured distance into point cloud data. The experimental results show that this 3D laser radar can accomplish the 3D object scanning and the environment 3D reconstruction task. In addition, the experiment result proves that the product of the camera focal length and the baseline length is the key factor to influence measurement accuracy.

• Korean Journal of Applied Biomechanics
• /
• v.33 no.4
• /
• pp.185-195
• /
• 2023

Objective: The purpose of this study is to improve TPI OnBaseU Power Test and golf performance by conducting CrossFit power training. Method: Three male golf players from University B participated in this study. They had 3 to 4 years of golf experience and participated in 8 weeks of CrossFit power training. Results: OnBaseU Power Test: There was a lot of improvement in Sit up throw (27.9%) and Chest pass (10.58%), but there was not much improvement in Baseline Toss (R5.9, L9.8%) and Vertical Jump (4.1%). Golf shot data: There was a very statistically significant difference in Club speed, Ball speed, and Total Length, which are related to speed, and there was no difference in Club path and Smash factor, which are related to accuracy and posture. Conclusion: CrossFit power training was effective in improving TPI OnBaseU Power Test and golf performance (Club speed, Ball speed, Total Length).

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.6
• /
• pp.3166-3187
• /
• 2017

For "generic" multivariate public key cryptography (MPKC) systems, experts believe that the Unbalanced Oil-Vinegar (UOV) scheme is a feasible signature scheme with good efficiency and acceptable security. In this paper, we address two problems that are to find inversion solution of quadratic multivariate equations and find another structure with some random Oil-Oil terms for UOV, then propose a novel signature scheme based on hyper-sphere (HS-Sign for short) which directly answers these two problems. HS-Sign is characterized by its adding Oil-Oil terms and more advantages compared to UOV. On the one side, HS-Sign is based on a new inversion algorithm from hyper-sphere over finite field, and is shown to be a more secure UOV-like scheme. More precisely, according to the security analysis, HS-Sign achieves higher security level, so that it has larger security parameters choice ranges. On the other side, HS-Sign is beneficial from both the key side and computing complexity under the same security level compared to many baseline schemes. To further support our view, we have implemented 5 different attack experiments for the security analysis and we make comparison of our new scheme and the baseline schemes with simulation programs so as to show the efficiencies. The results show that HS-Sign has exponential attack complexity and HS-Sign is competitive with other signature schemes in terms of the length of the message, length of the signature, size of the public key, size of the secret key, signing time and verification time.

• Physical Therapy Korea
• /
• v.29 no.1
• /
• pp.48-53
• /
• 2022

Background: The gastrocnemius tightness can easily occur. Gastrocnemius tightness results in gait disturbance. Thus, various interventions have been used to release a tight gastrocnemius muscle and improve gait performance. Moreover, focal muscle vibration (FMV) has recently been extensively researched in terms of tight muscle release and muscle performance. However, no study has investigated the effects of FMV application on medial gastrocnemius architectural changes. Objects: In this study, we aimed to investigate the effects of FMV on medial gastrocnemius architecture in persons with limited ankle dorsiflexion. Methods: Thirty one persons with <10° of passive ankle dorsiflexion participated in this study. We excluded persons with acute ankle injury within six months prior to study onset, a history of ankle fracture, leg length discrepancy greater than 2 cm, no history of neurological dysfunction, or trauma affecting the lower limb. The specifications of the FMV motor were as follows: a fixed frequency (fast wave: 150 Hz) and low amplitude (0.3-0.5 mm peak to peak) of vibration; the motor was used to release the medial gastrocnemius for 15 minutes. Each participant completed three trials for 10 days; a 30-second rest period was provided between each trial. Medial gastrocnemius architectural parameters [muscle thickness (MT), fiber bundle length (FBL), and pennation angle (PA)] were measured via ultrasonography. Results: MT significantly decreased after FMV application (p < 0.05). FBL significantly increased from its baseline value after FMV application (p < 0.05). PA significantly decreased from its baseline value after FMV application (p < 0.05). Conclusion: FMV application may be advantageous in reducing medial gastrocnemius excitability following a decrease in the amount of contractile tissue. Furthermore, FMV application can be used as a stretching method to alter medial gastrocnemius architecture.

• The Journal of the Acoustical Society of Korea
• /
• v.41 no.2
• /
• pp.122-129
• /
• 2022

In this paper, we propose a method to improve the performance of Transformer-based speech recognizers using data augmentation that locally adjusts the frame rate. First, the start time and length of the part to be augmented in the original voice data are randomly selected. Then, the frame rate of the selected part is changed to a new frame rate by using linear interpolation. Experimental results using the Wall Street Journal and LibriSpeech speech databases showed that the convergence time took longer than the baseline, but the recognition accuracy was improved in most cases. In order to further improve the performance, various parameters such as the length and the speed of the selected parts were optimized. The proposed method was shown to achieve relative performance improvement of 11.8 % and 14.9 % compared with the baseline in the Wall Street Journal and LibriSpeech speech databases, respectively.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.39 no.4
• /
• pp.251-261
• /
• 2003

To measure the GPS position accuracy and its distribution according to the length of the baseline, 30 minutes to 24 hours observations at the fixed location were conducted with two GPS receivers (Ll, 12 channels) on May 29 to June 2, 2002. The GPS data received at the reference station, the rover station and the ordinary times GPS observation station operated by the National Geography Institute in Korea were processed in kinematic and static post-processing methods with a post -processing software. The results obtained are summarized as follows: 1. The number of the satellite that could be observed continuously more than six hours was 16 and most of these satellites were positioned at east-west direction on May 31, 2002. The number of the satellite observed and the geometric dilution of precision (GDOP) determined by the average of every 10 minute for the day were 8 and 3.89, respectively. 2. Both the average GPS positions before and after post-processing were shifted (standalone: 1.17 m, post -processing: 0.43m) to the south and west. The twice distance root mean square (2drms) measured with standalone was 6.65m. The 2drms could be reduced to 33.8% (standard deviation 0=17.2) and 5.3% (0=2.2) of standalone by the kinematic and the static post-processing methods, respectively. 3. The relationship between the length of the baseline x (km) and the 2drms y (m) obtained by the static post-processing method was y=0.00l6x+0.006 $(R^2=0.87)$. In the case of the positioning with the static post-processing method using the GPS receiver, it was found that a positioning within 20cm 2drms was possible when the length of the baseline was less than 100km and the receiving time of the GPS is more than 30 minutes.

• Journal of ILASS-Korea
• /
• v.17 no.2
• /
• pp.94-99
• /
• 2012

The characteristics of spray behavior and injected amount were studied with two types of nozzles for using in a compression ignition engine with dual fuel technology for construction machines. A penetration length of spray tends to shorten due to a decrease of injected amount of a diesel fuel with dual fuel engine application. In order to ignite the gaseous fuel premixed with air during intake process, a diesel fuel, which was compression ignited, needs to penetrate somehow similar depth compared with the case of a diesel fuel-only-injection. In this work, a nozzle with reduced hole diameter and increased number of holes was tested and demonstrated that, compared to diesel 100% case, its penetration lengths are comparable to 74% and 79%, respectively, of those of 100% and 50% supply of a diesel fuel with the baseline nozzle that has four holes and 30.4% increased diameter. This will presumably enhancement the combustion in a dual fuel engine. A design suggestion was also made in this work to achieve similar penetration length of spray with diesel 100% case to prevent combustion from being deteriorated in a dual fuel engine.

• Structural Engineering and Mechanics
• /
• v.71 no.6
• /
• pp.751-761
• /
• 2019

Lamb wave technology is a promising technology in the field of structural health monitoring and can be applied in the detection and monitoring of defects in plate structures. Based on the reconstruction algorithm for the probabilistic inspection of damage (RAPID), a Lamb-based detection and evaluation method of through-thickness rectangular notches in metal plates was proposed in this study. The influences of through-thickness rectangular notch length and the angle between sensing path and notch length direction on signals were further explored through simulations and experiments. Then a damage index calculation method which focuses on both phase and amplitude difference between detected signals and baseline signals was proposed. Based on the damage index difference between two vertically crossed sensing paths which pass through the notch in a sensor network, the notch direction identification method was proposed. In addition, the notch length was determined based on the damage index distribution along sensing paths. The experimental results showed that the image reconstructed with the proposed method could reflect the information for the evaluation of notches.