• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.10
• /
• pp.4930-4951
• /
• 2018

The deep learning based speech enhancement has shown considerable success. However, it still suffers performance degradation under mismatch conditions. In this paper, an adaptation method is proposed to improve the performance under noise mismatch conditions. Firstly, we advise a noise aware training by supplying identity vectors (i-vectors) as parallel input features to adapt deep neural network (DNN) acoustic models with the target noise. Secondly, given a small amount of adaptation data, the noise-dependent DNN is obtained by using $L_2$ regularization from a noise-independent DNN, and forcing the estimated masks to be close to the unadapted condition. Finally, experiments were carried out on different noise and SNR conditions, and the proposed method has achieved significantly 0.1%-9.6% benefits of STOI, and provided consistent improvement in PESQ and segSNR against the baseline systems.

• Journal of the Korean Society of Industry Convergence
• /
• v.27 no.2_2
• /
• pp.343-356
• /
• 2024

Super app is an application that provides a variety of services in a unified interface within a single platform. With the acceleration of digital transformation, super apps are becoming more prevalent. This study aims to suggest service enhancement measures by analyzing the user review data before and after the transition to a super app. To this end, user review data from a payment-based super app(Shinhan Play) were collected and studied via topic modeling. Moreover, a matrix for assessing the importance and usefulness of topics is introduced, which relies on the eigenvector centrality of the inter-topic network obtained through topic modeling and the number of review recommendations. This allowed us to identify and categorize topics with high utility and impact. Prior to the transition, the factors contributing to user satisfaction included 'payment service,' 'additional service,' and 'improvement.' Following the transition, user satisfaction was associated with 'payment service' and 'integrated UX.' Conversely, dissatisfaction factors before the transition encompassed issues related to 'signup/installation,' 'payment error/response,' 'security authentication,' and 'security error.' Following the transition, user dissatisfaction arose from concerns regarding 'update/error response' and 'UX/UI.' The research results are expected to be used as a basis for establishing strategies to strengthen service competitiveness by making super app services more user-oriented.

• Phonetics and Speech Sciences
• /
• v.10 no.1
• /
• pp.33-38
• /
• 2018

Although deep neural network-based acoustic models have greatly improved the performance of automatic speech recognition (ASR), reverberation still degrades the performance of distant speech recognition in indoor environments. In this paper, we adopt the DenseNet, which has shown great performance results in image classification tasks, to improve the performance of reverberant speech recognition. The DenseNet enables the deep convolutional neural network (CNN) to be effectively trained by concatenating feature maps in each convolutional layer. In addition, we extend the concept of multi-resolution CNN to multi-resolution DenseNet for robust speech recognition in reverberant environments. We evaluate the performance of reverberant speech recognition on the single-channel ASR task in reverberant voice enhancement and recognition benchmark (REVERB) challenge 2014. According to the experimental results, the DenseNet-based acoustic models show better performance than do the conventional CNN-based ones, and the multi-resolution DenseNet provides additional performance improvement.

• Journal of electromagnetic engineering and science
• /
• v.18 no.3
• /
• pp.160-168
• /
• 2018

A wide spectrum of potential applications using substrate integrated waveguide (SIW) technologies in conjunction with air hole regions is introduced, and an efficient optimization methodology to cope with the multiple air hole effect in SIW applications is proposed. The methodology adopts a genetic algorithm to obtain optimum air hole dimensions for the specific propagation constant that can be accurately calculated using the recursive and closed form equations presented. The optimization results are evaluated by designing an SIW bandpass filter, and they show excellent performance. The optimization methodology using the proposed equations is effective in performance enhancement for the purposes of low loss and broadband SIW applications.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.5 no.4
• /
• pp.175-184
• /
• 2010

This paper designs a virtual USB drive for mobile devices which gives an illusion of a traditional USB flash memory drive and provides capacity-free storage space over IP network. The virtual USB drive operating with a S3C2410 hardware platform and embedded linux consists of USB device driver, an iSCSI-enabled network stack, and a seamless USB/iSCSI tunneling module. For performance enhancement, it additionally provides a kernel-level seamless USB/iSCSI tunneling module and data sharing with symbol references among kernel modules. Experiments reveal that the kernel-level implementation can improve the I/O performance up to 8 percentage, as compared with the user-level implementation.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.8 no.4
• /
• pp.25-33
• /
• 2008

The performance of power constrained cooperative multi-relay system with/without opportunistic transmission is considered in Rayleigh fading. The three power allocation methods are considered to maximize the system performance when the total network power is limited. It is analyzed that the opportunistic power allocation strategy has the best performance enhancement compared to the other power allocation strategies. The opportunistic relays increases with the total network power, which induce the higher diversity order of the opportunistic cooperative diversity, consequently improves the end-to-end outage probability.

• International Journal of Internet, Broadcasting and Communication
• /
• v.8 no.1
• /
• pp.64-68
• /
• 2016

These fields are integrated to visualize and finalize the proposed development, in simulation environment. SCADA (supervisory control and data acquisition) systems and distributed control systems (DCSs) are widely deployed in all over the world, which are designed to control the industrial infrastructures, in real ways. To supervise and control the various parts of designed systems; trends to require a deep knowledge to understand the overall functional needs of industries, which could be a big challenge. Industrial field devices (or network sensors) are usually distributed in many locations and are controlled from centralized site (or main control center); the communication provides various signs of security issues. To handle these issues, the research contribution will twofold: a method using cryptography is deployed in critical systems for security purposes and overall transmission is controlled from main controller site. At controller site, multimedia components are employed to control the overall transmission graphically, such as system communication, bytes flows, security embedded parameters and others, by the means of multimedia technology.

• Journal of the Korean Institute of Telematics and Electronics A
• /
• v.31A no.7
• /
• pp.1-16
• /
• 1994

In this paper, we propose a new multiple access scheme for VSAT(Very Small Aperture Terminal) data communications. The proposed scheme has better performance in terms of delay throughput than Controlled Multiaccess Scheme which has excellent performance. The proposed scheme use the spare reservation method, adaptation method to offered load, and fixed bandwidth reservation method. In this paper, we analyze the performance using simple queueing model and present the simulation results. When network traffic is very low, the new scheme and the controlled multiaccess scheme have almost the same access delay characteristic. As the network load becomes higher, the new scheme's access delay gain becomes larger. Futhermore, even when the network traffic is very high it is possible to access satellite link within the delay similar to one round-trip delay. In addition to access delay performance enhancement, the new scheme has facility that supports fixed bandwidth reservation. So it shows more enhanced performance in the environment that stream traffic is dominant such as in VSAT communication environment.

• The Journal of the Acoustical Society of Korea
• /
• v.17 no.4E
• /
• pp.3-10
• /
• 1998

This paper concerns the dynamical behavior, in probabilistic sense, of a feedforward neural network performing auto association for novelty. Networks of retinotopic topology having a one-to-one correspondence between and output units can be readily trained using back-propagation algorithm, to perform autoassociative mappings. A novelty filter is obtained by subtracting the network output from the input vector. Then the presentation of a "familiar" pattern tends to evoke a null response ; but any anomalous component is enhanced. Such a behavior exhibits a promising feature for enhancement of weak signals in additive noise. As an analysis of the novelty filtering, this paper shows that the probability density function of the weigh converges to Gaussian when the input time series is statistically characterized by nonsymmetrical probability density functions. After output units are locally linearized, the recursive relation for updating the weight of the neural network is converted into a first-order random differential equation. Based on this equation it is shown that the probability density function of the weight satisfies the Fokker-Planck equation. By solving the Fokker-Planck equation, it is found that the weight is Gaussian distributed with time dependent mean and variance.

• IEIE Transactions on Smart Processing and Computing
• /
• v.4 no.2
• /
• pp.110-114
• /
• 2015

This paper reports a novel object-detection technique in video sequences. The proposed algorithm consists of detection of objects in illuminated and occluded videos by using object features and a neural network technique. It consists of two functional modules: region-based object feature extraction and continuous detection of objects in video sequences with region features. This scheme is proposed as an enhancement of the Lowe's scale-invariant feature transform (SIFT) object detection method. This technique solved the high computation time problem of feature generation in the SIFT method. The improvement is achieved by region-based feature classification in the objects to be detected; optimal neural network-based feature reduction is presented in order to reduce the object region feature dataset with winner pixel estimation between the video frames of the video sequence. Simulation results show that the proposed scheme achieves better overall performance than other object detection techniques, and region-based feature detection is faster in comparison to other recent techniques.