• The Journal of the Acoustical Society of Korea
• /
• v.35 no.4
• /
• pp.303-309
• /
• 2016

This paper proposes a novel approach to the problem of audio signal clustering using a stacked autoencoder. The proposed stacked autoencoder learns an efficient representation for the input signal, enables clustering constituent signals with similar characteristics, and therefore the original sources can be separated based on the clustering results. STFT (Short-Time Fourier Transform) is performed to extract time-frequency spectrum, and rectangular windows at all the possible locations are used as input values to the autoencoder. The outputs at the middle, encoding layer, are used to cluster the rectangular windows and the original sources are separated by the Wiener filters derived from the clustering results. Source separation experiments were carried out in comparison to the conventional NMF (Non-negative Matrix Factorization), and the estimated sources by the proposed method well represent the characteristics of the orignal sources as shown in the time-frequency representation.

• Journal of Astronomy and Space Sciences
• /
• v.28 no.4
• /
• pp.267-271
• /
• 2011

I developed an enhanced correction method for Ricean bias which occurs in linear polarization measurement. Two known methods for Ricean bias correction are reviewed. In low signal-to-noise area, the method based on the mode of the equation gives better representation of the fractional polarization. But a caution should be given that the accurate estimation of noise level, i.e. ${\sigma}$ of the polarized flux, is important. The maximum likelihood method is better choice for high signal-to-noise area. I suggest a hybrid method which uses the mode of the equation at the low signal-to-noise area and takes the maximum likelihood method at the high signal-to-noise area. A modified correction coefficient for the mode solution is proposed. The impact on the depolarization measure analysis is discussed.

• Proceedings of the Korea Institute of Convergence Signal Processing
• /
• 2003.06a
• /
• pp.110-113
• /
• 2003

In this paper, the multi-channel high speed data acquisition system is implemented. This high speed signal processing system for 3-D image display is applicable to the manipulation of a medical image processing, multimedia data and various fields of digital image processing. In order to convert the analog signal into digital one, A/D conversion circuit is designed. PCI interface method is designed and implemented, which is capable of transmission a large amount of data to computer. In order to, especially, channel extendibility of images acquisition, bus communication method is selected. By using this bus method, we can interface each module effectively. In this paper, 32-channel A/D conversion and PCI interface system for 3-dimensional and real-time display of the retina image is developed. The 32-channel image acquisition system and high speed data transmission system developed in this paper is applicable to not only medical image processing as 3-D representation of retina image but also various fields of industrial image processing in which the multi-point realtime image acquisition system is needed.

• The Journal of the Acoustical Society of Korea
• /
• v.13 no.1E
• /
• pp.71-81
• /
• 1994

In this paper, we propose a pseudo-cepstral representation of line spectrum pair(LSP) frequencies and evaluate speech recognition performance with cepstral lift using the pseudo-cepstrum. The pseudo-cepstrum corresponding to LSP frequencies is derived by approxmating the relationship between LPC-cepstrum and LSP frequencies. Three cepstral liftering procedures are applied to the pseudo-cepstrum to improve the performance of speech recognition. They are the root-power-sums ligter, the general exponential lifter, and the bandpass lifter. Then, the liftered psedudo-cepstra are warped into a mel-frequency scale to obtain feature vectors for speech recognition. Among the three lifters, the general exponential lifter results in the best performance on speech recognition. When we use the proposed pseudo-cepstra feature vectors for recognizing noisy speech, the signal-to-noise ratio (SNR) improvement of about 5~10dB LSP is obtained.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 1994.10a
• /
• pp.670-675
• /
• 1994

Yhis paper describes a fault diagnosis simulation of the Real-Time Multiple Fault Dignosis System (RTMFDS) for forcasting faults in a system and deciding current machine state from signal information. Comparing with other diagnosis system for single fault,the system developed deals with multiple fault diagnosis,comprising two main parts. One is a remotesignal generating and transimission terminal and the other is a host system for fault diagnosis. Signal generator generate the random fault signal and the image information, and send this information to host. Host consists of various modules and agents such as Signal Processing Module(SPM) for sinal preprocessing, Performence Monotoring Module(PMM) for subsystem performance monitoring, Trigger Module(TM) for multi-triggering subsystem fault diagnosis, Subsystem Fault Diagnosis Agent(SFDA) for receiving trigger signal, formulating subsystem fault D\ulcornerB and initiating diagnosis, Fault Diagnosis Module(FDM) for simulating component fault with Hierarchical Artificial Neural Network (HANN), numerical models and Hofield network,Result Agent(RA) for receiving simulation result and sending to Treatment solver and Graphic Agent(GA). Each agent represents a separate process in UNIX operating system, information exchange and cooperation between agents was doen by IPC(Inter Process Communication : message queue, semaphore, signal, pipe). Numerical models are used to deseribe structure, function and behavior of total system, subsystems and their components. Hierarchical data structure for diagnosing the fault system is implemented by HANN. Signal generation and transmittion was performed on PC. As a host, SUN workstation with X-Windows(Motif)is used for graphic representation.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.5
• /
• pp.34-44
• /
• 2010

In this study, the algorithms for multichannel format conversion and robust representation of spatial sound information are proposed. In the spatial analysis, the directional information of sound source is estimated and sound sources are separated from stereo signal. In the spatial resynthesis, the multichannel matrixing with spatial repanning and post-scaling method are applied to represent a spatial sound. The conventional method about channel format conversion has the problem that the energy of sound source and the spatial information are not preserved in the desired channel format. Because the proposed method is designed in consideration of the target multichannel format and its resynthesized signal, the robust representation of spatial sound can be achieved in the multichannel format conversion.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.1
• /
• pp.53-76
• /
• 2020

The synchronous phasor measurement algorithm is the core content of the phasor measurement unit. This manuscript proposes a dynamic synchronous phasor measurement algorithm based on compressed sensing theory. First, a dynamic signal model based on the Taylor series was established. The dynamic power signal was preprocessed using a least mean square error adaptive filter to eliminate interference from noise and harmonic components. A Chirplet overcomplete dictionary was then designed to realize a sparse representation. A reduction of the signal dimension was next achieved using a Gaussian observation matrix. Finally, the improved orthogonal matching pursuit algorithm was used to realize the sparse decomposition of the signal to be detected, the amplitude and phase of the original power signal were estimated according to the best matching atomic parameters, and the total vector error index was used for an error evaluation. Chroma 61511 was used for the output of various signals, the simulation results of which show that the proposed algorithm cannot only effectively filter out interference signals, it also achieves a better dynamic response performance and stability compared with a traditional DFT algorithm and the improved DFT synchronous phasor measurement algorithm, and the phasor measurement accuracy of the signal is greatly improved. In practical applications, the hardware costs of the system can be further reduced.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.5
• /
• pp.1755-1777
• /
• 2022

The development of wireless communication technology has led to the underutilization of radio spectra. To address this limitation, an intelligent cognitive radio network was developed. Specific emitter identification (SEI) is a key technology in this network. However, in realistic non-cooperative scenarios, the system may detect signal classes beyond those in the training database, and only a few labeled signal samples are available for network training, both of which deteriorate identification performance. To overcome these challenges, a meta-learning-based open-set identification system is proposed for SEI. First, the received signals were pre-processed using bi-spectral analysis and a Radon transform to obtain signal representation vectors, which were then fed into an open-set SEI network. This network consisted of a deep feature extractor and an intrinsic feature memorizer that can detect signals of unknown classes and classify signals of different known classes. The training loss functions and the procedures of the open-set SEI network were then designed for parameter optimization. Considering the few-shot problems of open-set SEI, meta-training loss functions and meta-training procedures that require only a few labeled signal samples were further developed for open-set SEI network training. The experimental results demonstrate that this approach outperforms other state-of-the-art SEI methods in open-set scenarios. In addition, excellent open-set SEI performance was achieved using at least 50 training signal samples, and effective operation in low signal-to-noise ratio (SNR) environments was demonstrated.

• Journal of Positioning, Navigation, and Timing
• /
• v.8 no.4
• /
• pp.193-200
• /
• 2019

The auto-encoder network which is a good candidate to handle the modeling of the signal strength attenuation is designed for denoising and compensating the distortion of the received data. It provides a non-linear mapping function by iteratively learning the encoder and the decoder. The encoder is the non-linear mapping function, and the decoder demands accurate data reconstruction from the representation generated by the encoder. In addition, the adaptive network width which supports the automatic generation of new hidden nodes and pruning of inconsequential nodes is also implemented in the proposed algorithm for increasing the efficiency of the algorithm. Simulation results show that the proposed method can improve the neural network training surface to achieve the highest possible accuracy of the signal modeling compared with the conventional modeling method.

• Communications for Statistical Applications and Methods
• /
• v.22 no.2
• /
• pp.159-172
• /
• 2015

This paper considers the problem of reconstructing an underlying signal from noisy data. This paper presents a reconstruction method based on synchrosqueezed wavelet transform recently developed for multiscale representation. Synchrosqueezed wavelet transform based on continuous wavelet transform is efficient to estimate the instantaneous frequency of each component that consist of a signal and to reconstruct components. However, an objective selection method for the optimal number of intrinsic mode type functions is required. The proposed method is obtained by coupling the synchrosqueezed wavelet transform with cross-validation scheme. Simulation studies and musical instrument sounds are used to compare the empirical performance of the proposed method with existing methods.