• Smart Media Journal
• /
• v.12 no.5
• /
• pp.28-35
• /
• 2023

In this study, we propose a new light-weight model RoutingConvNet with fewer parameters to improve the applicability and practicality of speech emotion recognition. To reduce the number of learnable parameters, the proposed model connects bidirectional MFCCs on a channel-by-channel basis to learn long-term emotion dependence and extract contextual features. A light-weight deep CNN is constructed for low-level feature extraction, and self-attention is used to obtain information about channel and spatial signals in speech signals. In addition, we apply dynamic routing to improve the accuracy and construct a model that is robust to feature variations. The proposed model shows parameter reduction and accuracy improvement in the overall experiments of speech emotion datasets (EMO-DB, RAVDESS, and IEMOCAP), achieving 87.86%, 83.44%, and 66.06% accuracy respectively with about 156,000 parameters. In this study, we proposed a metric to calculate the trade-off between the number of parameters and accuracy for performance evaluation against light-weight.

• The Journal of the Acoustical Society of Korea
• /
• v.37 no.1
• /
• pp.73-81
• /
• 2018

The quality of side scan sonar images is determined by the frequency of a sonar. A side scan sonar with a low frequency creates low-quality images. One of the factors that lead to low quality is a high-level noise. The noise is occurred by the underwater environment such as equipment noise, signal interference and so on. In addition, in order to compensate for the transmission loss of sonar signals, the received signal is recovered by TVG (Time-Varied Gain), and consequently the side scan sonar images contain non-homogeneous noise which is opposite to optic images whose noise is assumed as homogeneous noise. In this paper, the SSCS (Structural Sparsity based Compressive Sensing) is proposed for removing non-homogeneous noise. The algorithm incorporates both local and non-local models in a structural feature domain so that it guarantees the sparsity and enhances the property of non-local self-similarity. Moreover, the non-local model is corrected in consideration of non-homogeneity of noises. Various experimental results show that the proposed algorithm is superior to existing method.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.8 no.3
• /
• pp.409-414
• /
• 2013

With the increase of multimedia information such as images, researches have been realized on how to extract the high-level semantic information from low-level visual information, and a variety of techniques have been proposed to generate this information automatically. However, most of these technologies extract the semantic information between single images, it's difficult to extract semantic information when a combination of multiple objects within the image. In this paper, we extract the visual features of objects within the image and training images stored in the DB and the features of each object are defined by measuring the similarity. Using ontology reasoner, each object feature within images infers the semantic information by positional relation and associative relation. With this, it's possible to infer semantic information between objects within images, we proposed a method for inferring more complicated and a variety of high-level semantic information.

• Asia-Pacific Journal of Business
• /
• v.11 no.3
• /
• pp.243-260
• /
• 2020

Purpose - This paper aims to investigate the properties of crude oil based derivative security (DLS) focusing on step-down type for comprehensive understanding of its risk. Design/methodology/approach - Kernel estimation is conducted to figure out statistical feature of the process of oil price. We simulate oil price paths based on kernel estimation results and derive probabilities of hitting the barrier and early redemption. Findings - The amount of issuance for crude oil based DLS is relatively low when base prices are below $40 while it is high when base prices are around $60 or $100, which is not consistent with kernel estimation results showing that oil futures prices tend to revert toward $46.14 and the mean-reverting speed is faster as oil price is lower. The analysis based on simulated oil price paths reveals that probability of early redemption is below 50% for DLS with high base prices and the ratio of the probability of early redemption to the probability of hitting barrier is remarkably low compared to the case for DLS with low base prices, as the chance of early redemption is deferred. Research implications or Originality - Empirical results imply that the level of the base price is a crucial factor of the risk for DLS, thus introducing a time-varying knock-in barrier, which is similar to adjust the base price, merits consideration to enhance protection for DLS investors.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.45 no.5
• /
• pp.192-200
• /
• 2008

Multimodal biometrics is a method for personal authentication and verification using more than two types of biometrics data. RBF based score-level fusion uses pattern recognition algorithm for multimodal biometrics, seeking the optimal decision boundary to classify score feature vectors each of which consists of matching scores obtained from several unimodal biometrics system for each sample. In this case, all matching scores are assumed to have the same reliability. However, in recent research it is reported that the quality of input sample affects the result of biometrics. Currently the matching scores having low reliability caused by low quality of samples are not currently considered for pattern recognition modelling in multimodal biometrics. To solve this problem, in this paper, we proposed the RBF based score-level fusion approach which employs quality information of input biometrics data to adjust decision boundary. As a result the proposed method with Qualify information showed better recognition performance than both the unimodal biometrics and the usual RBF based score-level fusion without using quality information.

• Proceedings of the KIPE Conference
• /
• 2003.07a
• /
• pp.430-433
• /
• 2003

IGBTS are widely used for the industrial inverters in the mid power range at low voltage (440V$\~$660V) application. Advantageous features of the device are simple gate drive and high speed switching capability. Due to these advantages the application of IGBTS is enlarging into the high power application. However, to increase the power handling capacity at lower input voltage level, the current rating in each bridge arm must be enlarged. Therefore the parallel operation of IGBT devices is essentially needed. This paper describes the feasible parallel structures of the power circuit for the mid & the high power inverters and introduces the important design condition for the parallel operation of IGBT devices. To verify feasibility of the IGBT parallel operation, the feature of several IGBT devices (EUPEC, SEMIKRON's IGBT) are investigated and the power stacks are implemented and tested with these devices. The experimental results show the good characteristics for the parallel operation of IGBTS.

• Journal of information and communication convergence engineering
• /
• v.12 no.3
• /
• pp.181-185
• /
• 2014

The lifetime of a solid-state drive (SSD) is limited because of the number of program and erase cycles allowed on its NAND flash blocks. Data cannot be overwritten in an SSD, leading to an out-of-place update every time the data are modified. This result in two copies of the data: the original copy and a modified copy. This phenomenon is known as write amplification and adversely affects the endurance of the memory. In this study, we address the issue of reducing wear leveling through efficient handling of write requests. This results in even wearing of all the blocks, thereby increasing the endurance period. The focus of our work is to logically divert the write requests, which are concentrated to limited blocks, to the less-worn blocks and then measure the maximum number of write requests that the memory can handle. A memory without the proposed algorithm wears out prematurely as compared to that with the algorithm. The main feature of the proposed algorithm is to delay out-of-place updates till the threshold is reached, which results in a low overhead. Further, the algorithm increases endurance by a factor of the threshold level multiplied by the number of blocks in the memory.

• The Journal of Korean Association of Computer Education
• /
• v.4 no.2
• /
• pp.105-114
• /
• 2001

This paper illustrates a calculus module which generates step-by-step solutions using J/Link that connects Java and Mathematica. Such a module provides intermediate and low level students with a practical environment where they can easily follow the solution paths on their own paces. The extra feature of this module depicts graphical images for a given function and its differentiated result to enhance the visual understandings of calculus concepts. Mathematica as a mathematical expert system that provides systematic mathematical knowledge to students with step-by-step solutions will be possibly extended to the tutorial or CMI development. The proposed module is implemented in a Java servlet that links to Mathematica FrontEnd. This approach results in adopting font systems to express two dimensional mathematical expressions in web documents as an alternative typesetting tool.

• Proceedings of the IEEK Conference
• /
• 1999.06a
• /
• pp.1063-1066
• /
• 1999

The delay of cardiac depolarization wave detection in the conventional pacemakers or AICD (automatic implantable cardioverter/ defibrillator, or ICD) has been overlooked. However, it is known that the delay may cause hemodynamic problems and may prevent the proper operation of a new automatic feature, automatic capture verification, that is to be appeared in the near-future devices. In order to reduce the effects of the delay, a delay prevention algorithm was developed and tested by applying three human electrograms. The algorithm set the sensing threshold just above the measured noise level to reduce the detection delay. It is found that the low threshold was able to reduce the delay by 20msec(average) In most cases. The implementation results showed reliability and efficacy of the algorithm, and the algorithm could be applicable to the existing hardware and software of the conventional pacemakers and AICD without any significant modifications.

• The Korean Journal of Cytopathology
• /
• v.6 no.2
• /
• pp.199-203
• /
• 1995

Chordoma is relatively uncommon tumor comprising $1\sim4%$ of primary malignant bone tumors, and believed to arise from the remnants of notochordal tissue. Because of its rare occurrence in the thoracic spine, we report a case of chordoma involving the thoracic spine. A 45-year-old male was sufferred from chest pain radiating to the back. Chest CT showed a well marginated, round huge mass with multiseptated enhancement at the thoracic spine from T5 to T8 level. After percutaneous needle aspiration, piecemeal resection of the tumor was done. On cytologic smears, two types of neoplastic cells were arranged in sheets and cords in mucinous background. One type of cells consisted of medium sized cells with pink cytoplasm and round nuclei. The other type had voluminous bubbly or clear cytoplasm divided by intracytoplasmic septae imparting a feathery or basket-like appearance. Histologically, the tumor showed lobulated feature divided by fibrous septae and the tumor cells were pink eosinophilic or physaliphorous in morphology. Immunohistochemically, the tumor cells revealed strong positivity for low(AE1) and high (AE3) molecular weight cytokeratins.