• KIISE Transactions on Computing Practices
• /
• v.21 no.8
• /
• pp.531-536
• /
• 2015

The speech recognition is an active research area in the human computer interface (HCI). The objective of this study is to control digital devices with voices. In addition, the mouse is used as a computer peripheral tool which is widely used and provided in graphical user interface (GUI) computing environments. In this paper, we propose a method of controlling the mouse with the real-time speech recognition function of a smartphone. The processing steps include extracting the core voice signal after receiving a proper length voice input with real time, to perform the quantization by using the learned code book after feature extracting with mel frequency cepstral coefficient (MFCC), and to finally recognize the corresponding vowel using hidden markov model (HMM). In addition a virtual mouse is operated by mapping each vowel to the mouse command. Finally, we show the various mouse operations on the desktop PC display with the implemented smartphone application.

• Journal of Internet Computing and Services
• /
• v.22 no.2
• /
• pp.59-68
• /
• 2021

As artificial intelligence(AI) technologies, which have made rapid growth recently, began to be applied to the marine environment such as ships, there have been active researches on the application of CNN-based models specialized for digital videos. In E-Navigation service, which is combined with various technologies to detect floating objects of clash risk to reduce human errors and prevent fires inside ships, real-time processing is of huge importance. More functions added, however, mean a need for high-performance processes, which raises prices and poses a cost burden on shipowners. This study thus set out to propose a method capable of processing information at a high rate while maintaining the accuracy by applying Quantization techniques of a deep learning model. First, videos were pre-processed fit for the detection of floating matters in the sea to ensure the efficient transmission of video data to the deep learning entry. Secondly, the quantization technique, one of lightweight techniques for a deep learning model, was applied to reduce the usage rate of memory and increase the processing speed. Finally, the proposed deep learning model to which video pre-processing and quantization were applied was applied to various embedded boards to measure its accuracy and processing speed and test its performance. The proposed method was able to reduce the usage of memory capacity four times and improve the processing speed about four to five times while maintaining the old accuracy of recognition.

• Journal of the Korean Institute of Telematics and Electronics D
• /
• v.34D no.8
• /
• pp.62-70
• /
• 1997

In this paper, we presetn a new method to display 3-D image modelled as a sum of 2-D sliced images by expanding the concept of the conventional 2-D optical correlator based on spatial matched filtr to the 3-D region. It is hsown that a arbitrary image can be constructed by an array of the correlation-peaks between pixel-to-pixel and propose the systhesis precedure of 3-D spatial-matched-fjilter using fresnel diffraction equation to display 3-D image in space. It is also shown that the quantization problem is severe when the systehsised filter function is displayed on the conventional LC-SLM. To overcome this problem, anonlinear quantizaton method using the sigmoid function is suggested, and this method can reduce the bias and the loss of high spatial-frequency information, and improve the diffraction efficiency. Finally, the suggested method is tested by computer simulation and then approved by some optical experiments with the conventional LC-SLM.

• Proceedings of the IEEK Conference
• /
• 2001.09a
• /
• pp.723-726
• /
• 2001

일반적으로 블록기반의 동영상 압축방식은 블록킹효과를 필연적으로 수반한다. 특히 저비트율의 동영상에서는 블록킹 효과가 다른 어떤 영상의 왜곡보다 많이 발생한다. 본 논문에서는 이러한 블록킹효과를 효율적으로 인간 시각체계에 적합하게 실시간으로 제거하는 후처리 알고리즘을 제안한다. 우선 복원된 영상에서 인간의 시각체계와 동영상의 특성에 따라 3가지의 모드로 분리하여 QP(quantization Parameter)에 따라 임계치를 변화함으로써 각 모드의 필터링 범위를 가변시켰다. 이후에 각 모드에 알맞은 일차원 및 적응형 필터링을 적용한다. 적용된 모드별 필터링은 과도한 블러링 현상을 방지하고 영상내의 실제 에지성분읓 보호하면서 효과적으로 블록킹효과를 제거한다. 본 논문에서 제안하는 알고리즘을 실험 영상에 적용하였을 경우에 주관적 화질 및 객관적 화질인 PSNR로 0.5dB 정도 향상되었다.

• The Journal of the Acoustical Society of Korea
• /
• v.17 no.7
• /
• pp.12-21
• /
• 1998

본 논문은 음성 부호화기인 MPMLQ(Multi Pulse Maximum Likehood Quantization)를 고정 소숫점 범용 DSP에 실시간으로 구현할 때 발생되는 계산량을 줄이기 위한 변형된 형태의 MPMLQdp 관한 것이다. MPMLQ는 음성 신호에서 선형 예측 계수와 피치 정보를 추출하고 남은 잔여 신호와 가장 유사한 여기 신호를 표현할 때 상관법을 이용 한다. 상관법은 DSP상에 구현할 때 계수 승산 오버플로우를 발생시킬 수 있으므로 연산후 항상 점검하여야 한다. 이것은 MPMLQ 구현시 전체 계산량의 많은 부분을 차지한다. 본 논 문은 이러한 문제점에 착안하여 계수 승산 오버플로우가 발생하지 않도록 입력 음성신호의 크기를 2비트 만큼 줄이고, 이로 인하여 같은 크기로 줄어든 잔여 신호를 고려하여 MPMLQ에서 여기신호의 크기를 표현하는 고정 코드북 이득표를 적절히 조절하였다. 실험 결과 변형된 MPMLQ의 SSNR은 0.040325dB(실험data기준) 향상되었으며, 계산량에 있어서 도 17.7%의 처리속도 향상되었다. 따라서 고정 소숫점 범용 DSP에 실시간 구현이 가능하였다.

• Journal of information and communication convergence engineering
• /
• v.6 no.1
• /
• pp.29-33
• /
• 2008

This paper presents the proposal and hardware design of JPEG baseline encoder. The JPEG encoder system consists of line buffer, 2-D DCT, quantization, entropy encoding, and packer. A fully pipelined scheme for JPEG encoder is adopted to speed-up an image compression. The proposed architecture was described in VHDL and synthesized in Xilinx ISE 7.1i and simulated by modelsim 6.1i. The results showed that the performance of the designed JPEG baseline encoder is higher than that demanded by real-time applications for $1024{\times}768$ image size. The designed JPEG encoder IP can be easily integrated into various application systems, such as scanner, PC camera, color FAX, and network camera, etc.

• 제어로봇시스템학회:학술대회논문집
• /
• 1988.10b
• /
• pp.832-836
• /
• 1988

"Motion Stereo" is quite useful for visual guidance of the robot, but most range finding algorithms of motion stereo have suffered from poor accuracy due to the quantization noise and measurement error. In this paper, 3-D position estimation and refinement scheme is proposed, and its performance is discussed. The main concept of the approach is to consider the entire frame sequence at the same time rather than to consider the sequence as a pair of images. The experiments using real images have been performed under following conditions : hand-held camera, static object. The result demonstrate that the proposed nonlinear least-square estimation scheme provides reliable and fairly accurate 3-D position information for vision-based position control of robot. of robot.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.25 no.1
• /
• pp.81-87
• /
• 1988

This paper proposes a new EBTC(extended block truncation coding) algorithm extended from the BTC for image compression. The EBTC has a capability to eliminate the defects of BTC, such as the deterioration of resolution or blocky effect,and to make a real-time processing like BTC. It shows better performances than the DPCM and the transform coding. Especially, it is a suitable coding method for the high quality picture transmission. It may be adequate to the system of transmission rate of 30-50 Mbits/sec. The picture quality has been scarecely degraded with a vector quantization to the EBTC output at the bit rate of 1.25 bits/pel. The bit rate of the scalar quantized EBTC method is 2.6-3.7 bits/pel.

• Proceedings of the IEEK Conference
• /
• 2003.11a
• /
• pp.83-86
• /
• 2003

In this paper, we propose a new hardware architecture for integer transform, quantizer operation of a new video coding standard H.264/JVT. We describe the algorithm to derive hardware architecture emphasizing the importance of area for low cost and low power consumption. The proposed architecture has been verified by PCI-interfaced emulation board using APEX-II Altera FPGA and also by ASIC synthesis using Samsung 0.18 ${\mu}{\textrm}{m}$ CMOS cell library. The ASIC synthesis result shows that the proposed hardware can operate at 100 MHz, processing more than 1, 300 QCIF video frames per second. The hardware is going to be used as a core module when implementing a complete H.264 video encoder/decoder ASIC for real-time multimedia application.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 1996.04a
• /
• pp.254-258
• /
• 1996

In this paper, we proposed realtime visual tracking system of moving object for 2D target using extended Kalman Filter Algorithm. A targeting marker are recongnized in each image frame and positions of targer object in each frame from a CCD camera while te targeting marker is attached to the tip of the SCARA robot hand. After the detection of a target coming into any position of the field-of-view, the target is tracked and always made to be located at the center of target window. Then, we can track the moving object which moved in inter-frames. The experimental results show the effectiveness of the Kalman filter algorithm for realtime tracking and estimated state value of filter, predicting the position of moving object to minimize an image processing area, and by reducing the effect by quantization noise of image