• 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.

• Communications of the Korean Mathematical Society
• /
• v.13 no.4
• /
• pp.791-800
• /
• 1998

We present a proof that, among all complex numbers, Duffin-Schaeffer's choice in the Neumann series expansion of the inverse of a frame operator has the best possible convergence rate.

• Journal of the Korean Institute of Telematics and Electronics A
• /
• v.33A no.10
• /
• pp.40-47
• /
• 1996

A method for computing effectiv ebandwidth of aggregated varable bit rate (VBR) moving picture experts group (MPEG) video traffic is proposed. To compute statistical characteristics of aggregated MPEG traffic, first we split input MPEG traffic into I, B, and P frame traffics and aggregate respective I, B, and P frame traffics according to the frame type. Second statisticsal characteristics of the aggregated MPEG traffic are obtained using those of aggregated I, B, and P frame traffics. The effective bandwidth of the aggregated I frame traffic is computed by using the gaussian bound. Using the modified equivalent capacity, we obtain the effective bandwidths of aggregated B and P frame traffics and then compute the effective bandwidth of the combined B and P frame traffic. Finally the effective bandwidth of the aggregated MPEG traffic is computed by adding the gaussian bound of the aggregated I frame traffic and modifed equivalent capacity of combined B and P frame traffic. Computer simulation shows that the proposed method estimates effective bandwidth of the aggregated MPEG traffic well.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.9 no.10
• /
• pp.1207-1214
• /
• 2014

The first frame of a GOP is encoded in intra mode which generates a larger number of bits. In addition, the first frame is used for the inter mode encoding of the following frames. Thus the encoding results of the intra frame affects the first frame as well as the following frames. Traditionally, the quantization parameter for an intra frame is determined only depending on the bpp not considering the characteristics of the intra frame. For accurate intra frame encoding, we should consider not only bpp but also the complexity of the video sequence and the output bandwidth. In this paper, we propose a real-time quantization model which is used to calculate the quantization parameter for an intra frame encoding based on the investigation on the characteristics of a GOP. It is shown by experimental results that the proposed quantization model captures the characteristics of an intra frame effectively and the proposed method for model parameters accurately estimates the real values.

• Proceedings of the IEEK Conference
• /
• 2003.07a
• /
• pp.53-56
• /
• 2003

In order to enhance the system capacity of multi rate IEEE 802.11 WLAN, we propose the transmission rate select method using the control frame in this paper. The transmission rate is selected dynamically based on the RSSI(Received Signal Strength Indicator) of received control frame. And we also propose the Enhanced NAV Update method for Fragmented MPDU burst transmission in multirate WLAN environment. This method can improve the system capacity by support the burst transmission of fragmented MPDU.

• Journal of Biomedical Engineering Research
• /
• v.24 no.3
• /
• pp.217-231
• /
• 2003

Coded excitation with complementary Golay sequences is an effective means to increase the SNR and penetration of ultrasound imaging. in which the two complementary binary codes are transmitted successively along each scan-line, reducing the imaging frame rate by half. This method suffers from low frame rate particularly when multiple transmit focusing is employed, since the frame rate will be further reduced in proportion to the number of focal zones. In this paper. a new ultrasound imaging technique based on simultaneous multiple transmit focusing using modified orthogonal Golay codes is proposed to improve lateral resolution with no accompanying decrease in the imaging frame rate, in which a pair of orthogonal Golay codes focused at two different focal depths are transmitted simultaneously. On receive, these modified orthogonal Golay codes are separately compressed into two short pulses and individually focused. These two focused beams are combined to form a frame of image with improved lateral resolution. The Golay codes were modified to improve the transmit power efficiency (TPE) for practical imaging. Computer simulations and experimental results show that the proposed method improves significantly the lateral resolution and penetration of ultrasound imaging compared with the conventional method.

• Transactions of the Korean Society of Automotive Engineers
• /
• v.11 no.6
• /
• pp.197-204
• /
• 2003

This paper develops a computer model for estimating the handling of a cabover type large-sized truck. The truck is composed of front and rear suspension systems, a frame, a cab, and ten tires. The computer model is developed using ADAMS. A shock absorber, a rubber bush, and a leaf spring aunt a lot on the dynamic characteristic of the vehicle. Their stiffness and damping coefficient are measured and used as input data of the computer model. Leaf springs in the front and rear suspension systems are modeled by dividing them three links and joining them with joints. To improve the reliability of the developed computer model, the frame is considered as a flexible body. Thus, the frame is modeled by finite elements using MSC/PATRAN. A mode analysis is performed with the frame model using MSC/NASTRAN in order to link the frame model to the computer model. To verify the reliability of the developed computer model, a double lane change test is performed with an actual vehicle. In the double lane change, lateral acceleration, yaw rate, and roll angle are measured. Those test results are compared with the simulation results.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.9
• /
• pp.2027-2034
• /
• 1997

Frame relay is a fast packet switching technology that performs relaying and multiplexing frames with variable lengths over a wide area link at the T1 or E1 speed, by elminating error and flow control in the network. In frame relay networks, congestion control is typically performed through the rate enforcement with a discard eligibility (DE) bit, and the explicit negative feedback meachanisms using explicit congetion notification bits. In this paper, we consider the congestiong control scheme using the rate enforcement mechanism with DE bit for frame relay network. Assuming that each frame with exponentially distributed length arrives according to the Poission fashion, we can treat the frame relay switch as an M/M/1/K priority queueing system with pushout basis. We analyze and present the blocking probabilities and waiting time distributions of frames.

• Transactions of the Korean Society for Noise and Vibration Engineering
• /
• v.15 no.6 s.99
• /
• pp.731-737
• /
• 2005

This paper evaluates dynamic performance of a cab over type large-sized truck for estimating the effects of frame's torsional characteristics using a computer model. The computer model considers two mounting methods of frame, flange mounting and web mounting. Frame is modeled by finite elements using MSC/NASTRAN In order to consider the flexibility of frame. The torsional test of the frame is conducted In order to validate the modeled finite element model. A load cell is used to measure the load applied to the frame. An angle sensor is used to measure the torsional angle. An actuator is used to apply a load to the frame. To estimate the effects of frame's torsional characteristics on dynamic performance, simulations are performed with the flange mounting and web mounting frame. Simulation results show that the web mounting frame's variations of roll angle, lateral acceleration, and yaw rate are larger than the flange mounting frame's variations, especially in the high velocity and the second part of the double lane course.

• Journal of Broadcast Engineering
• /
• v.8 no.4
• /
• pp.501-504
• /
• 2003

In this out watermarking method, a key is determined by quantizing the maximum motion difference between frames. We have a problem the key value for embedding and detection are different in 1 to 3% frames of all frames. This problem can be easily solved by using a new key according to bit error rate of the extracted watermark. Since the watermark is embedded in each frame nth different keys and detected In all the frames, out method is resistant against attacks such as the frame averaging and frame drop.