• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.66 no.9
• /
• pp.1364-1372
• /
• 2017

In this paper, a problem of event-triggered model predictive control is investigated for continuous-time Takagi-Sugeno (T-S) fuzzy systems with input quantization. To efficiently utilize network resources, event-trigger is employed, which transmits limited signals satisfying the condition that the measurement of errors is over the ratio of a certain level. Considering sampling and quantization, continuous Takagi-Sugeno (T-S) fuzzy systems are regarded as a sector bounded continuous-time T-S fuzzy systems with input delay. Then, a model predictive controller (MPC) based on parallel distributed compensation (PDC) is designed to optimally stabilize the closed loop systems. The proposed MPC optimize the objective function over infinite horizon, which can be easily calculated and implemented solving linear matrix inequalities (LMIs) for every event-triggered time. The validity and effectiveness are shown that the event triggered MPC can stabilize well the systems with even smaller average sampling rate and limited actuator signal guaranteeing optimal performances through the numerical example.

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.35C no.2
• /
• pp.56-64
• /
• 1998

Multi-layered perceptrons that are a nonlinear neural network model, have been widely used for various applications mainly thanks to good function approximation capability for nonlinear fuctions. However, for digital hardware implementation of the multi-layere perceptrons, the quantization scheme using "look-up tables (LUTs)" is commonly employed to handle nonlinear signmoid activation functions in the neworks, and thus requires large amount of storage to prevent unacceptable quantization errors. This paper is concerned with a new effective methodology for digital hardware implementation of multi-layered perceptrons, and proposes a "piecewise affine approximation" method in which input domain is divided into (small number of) sub-intervals and nonlinear sigmoid function is linearly approximated within each sub-interval. Using the proposed method, we develop an expression and an error backpropagation type learning algorithm for a multi-layered perceptron, and compare the performance with the quantization method through Monte Carlo simulations on XOR problems. Simulation results show that, in terms of learning convergece, the proposed method with a small number of sub-intervals significantly outperforms the quantization method with a very large storage requirement. We expect from these results that the proposed method can be utilized in digital system implementation to significantly reduce the storage requirement, quantization error, and learning time of the quantization method.quantization method.

• Proceedings of the IEEK Conference
• /
• 2000.09a
• /
• pp.667-670
• /
• 2000

This paper provides a heuristic theory for modeling and analysis of vector quantization effects in 2-dimensional subband filter banks. This model is used as the basis for optimal filter bank design. The scalar non-linear gain-plus-additive noise quantization model can be used to represent each vector quantizer in 2-band subband codec. The validity and accuracy and of this analytic model is confirmed by comparing the calculated model quantization errors with actual simulation of the optimum LBG vector quantizer. Numerical design examples for the optimum separable paraunitary filter banks are suggested in this paper.

• The Journal of Korea Robotics Society
• /
• v.17 no.2
• /
• pp.133-141
• /
• 2022

This paper introduces model compression algorithms which make a deep neural network smaller and faster for embedded systems. The model compression algorithms can be largely categorized into pruning, quantization and knowledge distillation. In this study, gradual pruning, quantization aware training, and knowledge distillation which learns the activation boundary in the hidden layer of the teacher neural network are integrated. As a large deep neural network is compressed and accelerated by these algorithms, embedded computing boards can run the deep neural network much faster with less memory usage while preserving the reasonable accuracy. To evaluate the performance of the compressed neural networks, we evaluate the size, latency and accuracy of the deep neural network, DenseNet201, for image classification with CIFAR-10 dataset on the NVIDIA Jetson Xavier.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.45 no.3
• /
• pp.93-102
• /
• 2008

In this paper, we propose a rate control algorithm for constant quality real time video coding. To achieve constant quality, previous algorithm exploit mean absolute of difference(MAD) as measure of frame complexity. However, if scene is abruptly changed or if quantization parameter is not constant, encoder produces various output bits with same MAD. Therefore we know that MAD does not appropriately reflect characteristic of frame. To solve this problem, we exploit model parameter as measure of frame complexity. Because model parameter means slope between output bits and MAD, it reflects correctly complexity of frame. And because previous model, R-MAD model, is not considered quantization parameter, as quantization parameter increases or decreases, model parameter of frame also vary. So model parameter obtained using previous model cannot reflect internal characteristic of video. We solve this problem using proposed model, which is considered quantization parameter. Experiment results show that our algorithm provide better performance, in terms of quality smoothness than previous algorithm. Especially, when scene is abruptly changed, our algorithm alleviates quality drop.

• The Journal of the Acoustical Society of Korea
• /
• v.16 no.1E
• /
• pp.41-48
• /
• 1997

The very purpose of subband codec is the attainment of data rate compression through the use of quantizer and optimum bit allocation for each decimated signal. Yet the question of floating-point quantization effects in subband codec has received scant attention. There has been no direct focus on the analysis of quantization errors, nor on design with quantization errors embedded explicitly in the criterion. This paper provides a rigorous theory for the modelling, analysis and optimum design of the general M-band subband codec in the presence of the floating-point quantization noise. The floating-point quantizers are embedded into the codec structure by its equivalent multiplicative noise model. We then decompose the analysis and synthesis subband filter banks of the codec into the polyphase form and construct an equivalent time-invariant structure to compute exact expression for the mean square quantization error in the reconstructed an equivalent time-invariant structure to compute exact expression for the mean square quantization error in the reconstructed output. The optimum design criteria of the subband codec is given to the design of the analysis/synthesis filter bank and the floating-point quantizer to minimize the output mean square error. Specific optimum design examples are developed with two types of filter of filter banks-orthonormal and biorthogonal filter bank, along with their perpormance analysis.

• Journal of Broadcast Engineering
• /
• v.19 no.3
• /
• pp.329-341
• /
• 2014

Kernel-based mean-shift object tracking has gained more interests nowadays, with the aid of its feasibility of reliable real-time implementation of object tracking. This algorithm calculates the best mean-shift vector based on the color histogram similarity between target model and target candidate models, where the color histograms are usually produced after uniform color-space quantization for the implementation of real-time tracker. However, when the image of target model has a reduced contrast, such uniform quantization produces the histogram model having large values only for a few histogram bins, resulting in a reduced accuracy of similarity comparison. To solve this problem, a non-uniform quantization algorithm has been proposed, but it is hard to apply to real-time tracking applications due to its high complexity. Therefore, this paper proposes a fast non-uniform color-space quantization method using the histogram equalization, providing an adjusted histogram distribution such that the bins of target model histogram have as many meaningful values as possible. Using the proposed method, the number of bins involved in similarity comparison has been increased, resulting in an enhanced accuracy of the proposed mean-shift tracker. Simulations with various test videos demonstrate the proposed algorithm provides similar or better tracking results to the previous non-uniform quantization scheme with significantly reduced computation complexity.

• The Journal of the Acoustical Society of Korea
• /
• v.20 no.7
• /
• pp.76-82
• /
• 2001

In this paper, a novel quantization method for wideband signal codec with 7 kHz bandwidth is proposed. In the transform-based wideband signal codecs, the signal is transformed to frequency domain and the spectral coefficients in each frequency band are quantized based on human perceptual model, followed by Huffman coding. However, the property of each band varies with frequency, and the codec has poor performance when all bands are quantized with the same method. Therefore, a selective quantization method is proposed, which analyzes the band property and selects the quantization domain between frequency domain and time domain based on the quantization efficiency. It is confirmed that the proposed method has better performance than the quantizer of G722.1 codec.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.21 no.1
• /
• pp.93-98
• /
• 2021

Due to the development of deep learning and AI, the scale of the model has grown, and it has been integrated into other fields to blend into our lives. However, in environments with limited resources such as embedded devices, it is exist difficult to apply the model and problems such as power shortages. To solve this, lightweight methods such as clouding or offloading technologies, reducing the number of parameters in the model, or optimising calculations are proposed. In this paper, quantization of learned models is applied to ONNX models used in various framework interchange formats, neural network structure and inference performance are compared with existing models, and various module methods for quantization are analyzed. Experiments show that the size of weight parameter is compressed and the inference time is more optimized than before compared to the original model.

• ETRI Journal
• /
• v.38 no.1
• /
• pp.185-194
• /
• 2016

In this paper, a novel rate control scheme for low delay video coding of High Efficiency Video Coding (HEVC) is proposed. The proposed scheme is developed by considering a new temporal prediction structure of HEVC. In the proposed scheme, the relationship between bit rate and quantization step is exploited firstly to formulate an accurate quadratic rate-quantization (R-Q) model. Secondly, a method of determining the quantization parameters (QPs) for the first frames within a group of pictures is proposed. Thirdly, an accurate frame-level bit allocation method is proposed for HEVC. Finally, based on the proposed R-Q model and the target bit allocated for the frame, the QPs are predicted for coding tree units by using rate-distortion (R-D) optimization. We compare our scheme against that of three other state-of-the-art rate control schemes. Experimental results show that the proposed rate control scheme can increase the Bjøntegaard delta peak signal-to-noise ratio by 0.65 dB and 0.09 dB on average compared with the JCTVC-I0094 and JCTVC-M0036 schemes, respectively, both of which have been implemented in an HEVC test model encoder; furthermore, the proposed scheme achieves a similar R-D performance to Wang's scheme, as well as obtaining the smallest bit rate mismatch error of all the schemes.