• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.87-90
• /
• 2020

GPU의 발전과 함께 성장한 딥러닝(Deep Learning)은 영상 분류 문제에서 최고의 성능을 보이고 있다. 그러나 합성곱 신경망 기반의 모델을 깊게 쌓음에 따라 신경망의 표현력이 좋아짐과 동시에 때로는 학습이 잘되지 않고 성능이 저하되는 등의 부작용도 등장했다. 성능 향상을 방해하는 주요 요인 중 하나는, 차원감소 목적에 따라 필연적으로 정보 손실을 겪어야 하는 풀링 계층에 있다. 따라서 특성맵(Feature map)의 차원감소를 통해 얻게 되는 비용적 이득과 모델의 분류 성능 사이의 취사선택(Trade-off)이 존재한다. 그리고 이로부터 자유로워지기 위한 다양한 연구와 기법이 존재하는데 Spectral Pooling도 이 중 하나이다. 본 논문에서는 이산 푸리에 변환(Discrete Fourier Transform, DFT)을 이용한 Spectral Pooling에 대한 소개와, 해당 풀링의 성질을 통상적으로 사용되고 있는 Max Pooling과의 성능 비교를 통해 분석한다. 또한 영상 내 고주파수 부분에서 특히 더 강건하지 못하다는 맥스 풀링의 고질적인 문제점을, Spectral Pooling과의 하이브리드(Hybrid) 구조를 통해 어떻게 극복해나갈 것인지 그 가능성을 중심으로 실험을 수행했다.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.44 no.6
• /
• pp.29-34
• /
• 2007

It is important to analyse the frequency spectrum for the measurement of modulated signal, distortion, and noise. The frequency spectrum analysis is performed by the execution of Radix-2 DIT DFT i.e. FFT algorithm. The discrete input signal converted by A/D converter from the input signal in time domain is mathematically transformed to the frequency spectrum by FFT algorithm. In this study, we design the digital spectrum analyser by the hardware based on the TMS320F2812 DSP and AD9244 converter, and by the software based on the C28x S/W modules. We can timely analyse the frequency spectrum in mobile communication system by the digital frequency analyser based on the high performance DSP and S/W modules. This real-time analysing capability is the important performance in the internet-based mobile communication server system.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.27 no.12
• /
• pp.1878-1888
• /
• 1990

In this paper, the input-output interconnection method of the multi-valued signal processing circuit using perfect Shuffle technique and Kronecker product is discussed. Using this method, the design method of circuit of the multi-valued Reed-Muller expansions(MRME) to be used the multi-valued signal processing on finite field GF(p**m) is presented. The proposed input-output interconnection method is shown that the matrix transform is efficient and that the module structure is easy. The circuit design of MRME on FG(p**m) is realized following as` 1) contructing the baisc gates on GF(3) by CMOS T gate, 2) designing the basic cells to be implemented the transform and inverse transform matrix of MRME using these basic gates, 3) interconnecting these cells by the input-output interconnecting method of the multivalued signal processing circuits. Also, the circuit design of the multi-valued signal processing function on GF(3\ulcorner similar to Winograd algorithm of 3x3 array of DFT (discrete fourier transform) is realized by interconnection of Perfect Shuffle technique and Kronecker product. The presented multi-valued signal processing circuits that are simple and regular for wire routing and posses the properties of concurrency and modularity are suitable for VLSI.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.49 no.10
• /
• pp.69-80
• /
• 2012

In this paper, we studied the time-domain measurement and analysis techniques using a network analyzer for characterization UWB antenna link radiating impulse signal. For this purpose, we developed the CZT(Chirp z-Transform) algorithm which has characterized zoom-in function and transformed the acquired data from network analyzer to time domain format. Using the CZT algorithm, we proves that it would be better efficient and more faster than the DFT for analyzing the waveform and also be able to zoom-in the arbitrary region.

• International Journal of Highway Engineering
• /
• v.16 no.5
• /
• pp.59-66
• /
• 2014

PURPOSES : This study is to predict the Sound Pressure Level(SPL) obtained from the Noble Close ProXimity(NCPX) method by using the Extended Kalman Filter Algorithm employing the taylor series and Linear Regression Analysis based on the least square method. The objective of utilizing EKF Algorithm is to consider stochastically the effect of error because the Regression analysis is not the method for the statical approach. METHODS : For measuring the friction noise between the surface and vehicle's tire, NCPX method was used. With NCPX method, SPL can be obtained using the frequency analysis such as Discrete Fourier Transform(DFT), Fast Fourier Transform(FFT) and Constant Percentage Bandwidth(CPB) Analysis. In this research, CPB analysis was only conducted for deriving A-weighted SPL from the sound power level in terms of frequencies. EKF Algorithm and Regression analysis were performed for estimating the SPL regarding the vehicle velocities. RESULTS : The study has shown that the results related to the coefficient of determination and RMSE from EKF Algorithm have been improved by comparing to Regression analysis. CONCLUSIONS : The more the vehicle is fast, the more the SPL must be high. But in the results of EKF Algorithm, SPLs are irregular. The reason of that is the EKF algorithm can be reflected by the error covariance from the measurements.

• International Journal of Highway Engineering
• /
• v.15 no.6
• /
• pp.93-101
• /
• 2013

PURPOSES : The rumble strip installed at the highway near the tollgate has the purpose to reduce the vehicle velocity or prevent sleepiness by awakening people to the danger. These rumble strip has different vibration decibel from the rumble strip shapes, resulting in some fatigue damage to human because a driver suffers from a lot of stress and displeasure. In this connection, the objective of this paper is to analyze the vibration decibel perceived by a driver in the vehicle under some conditions. METHODS : The vibration decibel conveyed from the tire can be analyzed. The frequency analysis methods were used according to DFT (Discrete Fourier Transform) analysis, FFT (Fast Fourier Transform) analysis, CPB (Constant Percentage Bandwidth) analysis. But the frequency analysis method in this paper is the 1/24 OCT(Octave) band analysis because of the convenience of the analysis and the overall vibration amplitude along the frequency. RESULTS : By using the results of the vibration decibel after analyzing the 1/24 OCT band analysis, these results can be compared from some conditions (e.g., rumble strip shape, uniform velocity of a vehicle, road condition, mass of a vehicle). As a result, the numerical values of decibel are not directly proportional to the vehicle velocity. CONCLUSIONS : At the condition that a vehicle is passing by the rumble strip, the value of a vibration decibel at the rumble strip of the cylinder shape is smaller than the rumble strip of rectangular shape regardless of the rumble strip depth and width. At the mass condition, the more a vehicle is massive, the more the vibration decibel increases. At the road condition, the vibration decibel at the wet road is smaller than the value at dry road condition.

• International Journal of Highway Engineering
• /
• v.15 no.6
• /
• pp.79-91
• /
• 2013

PURPOSES : The methods of measuring the sound from the noise source are Pass-by method and NCPX (Noble Close Proximity) method. These measuring methods were used to determine the linkage of TAPL (Total Acoustic Pressure Level) and SPL (Sound Pressure Level) in terms of frequencies. METHODS : The frequency analysis methods are DFT (Discrete Fourier Transform) and FFT (Fast Fourier Transform), CPB (Constant Percentage Bandwidth). The CPB analysis was used in this study, based on the 1/3 octave band option configured for the frequency analysis. Furthermore, the regression analysis was used at the condition related to the sound attenuation effect. The MPE (Mean Percentage Error) and RMSE (Root Mean Squared Error) were utilized for calculating the error. RESULTS : From the results of the CPB frequency analysis, the predicted SPL along the frequency has 99.1% maximum precision with the measured SPL, resulting in roughly 1 dB(A) error. The TAPL results have precision by 99.37% with the measured TAPL. The predicted TAPL results at this study by using the SPL prediction model along the frequency have the maximum precision of 98.37% with the vehicle velocity. CONCLUSIONS : The Predicted SPL model along the frequency and the TAPL result by using the predicted SPL model have a high level of accuracy through this study. But the vehicle velocity-TAPL prediction model from the previous study by using the log regression analysis cannot be consistent with the TAPL result by using the predicted SPL model.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.37 no.4
• /
• pp.65-71
• /
• 2000

Finite wordlength effect of the recursive implementation of SDFT(sliding-DFT) is analytically derived in this paper. Representation errors of the twiddle coefficients and the data registers are the two major causes of the spectral errors in the recursive implementation. The noise-to-signal ratio is analytically derived in terms of the coefficients wordlength, the data registers wordlength, and the DFT's block-length used in the computation Error dynamic equation is obtained from the recursive DFT and the probabilistic models for the coefficients error and the round-off error are introduced for the NSR derivation, The result of the NSR derivation is verified with the simulation data.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.16 no.10 s.101
• /
• pp.1028-1035
• /
• 2005

OFDM system has high PAPR(Peak-to-Average-Power Ratio) problem. In this paper, we present a low complexity phase weighting method to reduce the computational quantity so that we can cut down the processing time of SPW method. Proposed method is derived from the DFT property of periodical sequences by which PAPR can be reduced efficiently. The simulation results show the same PAPR reduction efficiency of proposed method in comparison with conventional methods. It can reduce 2.15 dB of PAPR with two phase factors and 3.95 dB of PAPR with four phase factors. The computation analysis shows significant improvement in the low complexity phase weighting method.

• Journal of Advanced Marine Engineering and Technology
• /
• v.40 no.9
• /
• pp.818-823
• /
• 2016

Millimeter wave (mm-wave) communication systems provide high data rates owing to the large bandwidths available at mm-wave frequencies. Recently, analogue and digital combined beamforming, namely "hybrid beamforming" has drawn attentions owing to its ability to realize the required link margins in mm-wave systems. Taking into account the radio frequency (RF) hardware limitations, such as the analogue phase shifter gain constraint and the low resolution of the phase controller, we introduce an uplink hybrid beamforming system that includes discrete Fourier transform (DFT) based "fixed" analogue beamforming. We adopt a zero-forcing (ZF) multiple-input multiple-output (MIMO) equalizer to eliminate the uplink inter-user interferences. Moreover, to improve the sum-rate performances, we propose a transmit beam selection algorithm which makes the uplink effective channels, i.e., the beamformed channels, become near orthogonal. The effectiveness of the proposed beam selection algorithm was verified through numerical simulations.