Uipil Chong

Facial Expression Recognition Using Local Transitional Pattern on Gabor Filtered Facial Images

Abstract Automatic facial expression analysis plays a major role in catching emotional state of a human being and can be effectively used in the field of human-computer interaction. Using an effective facial feature is the most critical part for a successful facial expression recognition system. This paper proposes a novel approach in pursuit of recognizing facial expression where facial feature is represented by a hybrid of Gabor wavelet transform of an image and local transitional patterncode. Expression images are classified into prototype expressions via support vector machine with different kernels. Experimental results using Cohn-Kanade expression database is compared with other methods to demonstrate the superiority of the proposed approach which successfully identifies more than 95% of facial expressions correctly.

Fault diagnosis of induction motors utilizing local binary pattern-based texture analysis

Fault diagnosis of induction motors in the practical industrial fields is always a challenging task due to the difficulty that lies in exact identification of fault signatures at various motor operating conditions in the presence of background noise produced by other mechanical subsystems. Several signal processing approaches have been adopted so far to mitigate the effect of this background noise in the acquired sensor signal so that fault-related features can be extracted effectively. Addressing this issue, this paper proposes a new approach for fault diagnosis of induction motors utilizing two-dimensional texture analysis based on local binary patterns (LBPs). Firstly, time domain vibration signals acquired from the operating motor are converted into two-dimensional gray-scale images. Then, discriminating texture features are extracted from these images employing LBP operator. These local feature descriptors are later utilized by multi-class support vector machine to identify faults of induction motors. The efficient texture analysis capability as well as the gray-scale invariance property of the LBP operators enables the proposed system to achieve impressive diagnostic performance even in the presence of high background noise. Comparative analysis reveals that LBP8,1 is the most suitable texture analysis operator for the proposed system due to its perfect classification performance along with the lowest degree of computational complexity.

Improvement in Moving Target Detection Based on Hough Transform and Wavelet

ABSTRACT The most fundamental problem in radar signal is the detection of a moving object or a physical phenomenon. This problem became more challenging to the signal processing and radar communities under severe interference. In order to effectively detect moving targets in strong noise conditions, we proposed a new method based on wavelet and Hough transform for multiple moving targets detection. First, the noise is removed from the received signal by the wavelet technique. We propose an improved wavelet threshold function. The new threshold function not only reduces constant error of soft thresholding but also improves the discontinuity of hard thresholding. Second, the image processing technique of the Hough transform is used. The benefits of this method include improved performance of detection and implementation flexibility, which indicates the effectiveness and robustness of the algorithm.

Front-back confusion resolution in three-dimensional sound localization using databases built with a dummy head

Sound localization plays an important role in everyday life. It helps us to separate sounds coming from different sources and thus to acquire acoustic information. This paper describes an algorithm for localizing the position of a sound source, as recorded by dummy head microphones. The recorded signals are considered to be basic, random signals within an imaginary round room. The goal of this research is to localize random signals produced from different positions using information about basic signals. The method used is based on the identification of similarities between basic and random signals. It includes an interaural time difference comparison at the beginning, and continues with further analysis of the differences in signal spectrums. One of the main issues arising in sound localization is the problem of front-back confusion, and this paper shows how it was resolved by the use of reference signals.

Identification of significant intrinsic mode functions for the diagnosis of induction motor fault

For the analysis of non-stationary signals generated by a non-linear process like fault of an induction motor, empirical mode decomposition (EMD) is the best choice as it decomposes the signal into its natural oscillatory modes known as intrinsic mode functions (IMFs). However, some of these oscillatory modes obtained from a fault signal are not significant as they do not bear any fault signature and can cause misclassification of the fault instance. To solve this issue, a novel IMF selection algorithm is proposed in this work.

Noise reduction of continuous wave radar and pulse radar using matched filter and wavelets

This paper analyzes noise reduction using matched filter and wavelet transform in the signals of continuous wave radar and pulse radar. The denoising application of wavelets has been used in spectrum cleaning of atmospheric radar signals. Matched filter has a strong anti-noise ability; it can also achieve accurate pulse compression in a very noisy environment. This paper analyzes the algorithms of matched filter and wavelets that are used in radar signal processing to reduce the noise. The simulation results indicate that matched filter has a strong anti-noise ability for pulse radar and wavelet for continuous wave radar.

Design of a Power Factor Measurement System for Nonlinear Load

This paper introduces and develops an efficient method for measuring power factor (PF) and its nature under nonlinear load current situations. The method is based on generating a pulse width modulated signal whose width correlates to the value of PF. This signal can then be employed as a feedback signal for controlling PF related power quantities in a system. This method has the advantages of its simple implementation, less computational complexity, and its allowable error of less than 4[%], which is justified by the computer simulation results.

Real-time sound synthesis of plucked string instruments using a data parallel architecture

Recent advances in physics-based sound synthesis have offered huge potential possibilities for the creation of new musical instruments. Despite that research on physics-based sound synthesis is going on for almost three decades, its higher computational complexity has limited its use in real-time applications. Conventional serial computation is inadequate for handling the physics-based sound synthesis of most instruments. To yield computation time compatible with real-time performance, we introduce a parallel approach to the physics-based sound synthesis. In this paper, with a parallel processing engine we implemented the physical modeling for one of traditional Korean plucked string instruments, called Gayageum, which has 12 silk strings. Analysis and simulation results suggest that our parallel approach has the potential to support the real-time sound synthesis of the Gayageum instrument. Moreover, our parallel approach outperforms todays DSPs in terms of performance and energy efficiency.

A High-capacity Audio Watermarking Scheme in the Time Domain Based on Multiple Embedding

Abstract This paper introduces a new multiple embedding technique for the design of a time domain audio watermarking scheme. The concept of multiple embedding is realized by exploiting the properties of Polar coordinate system. In fact, more than one marking space is obtained from the host audio where watermarks are embedded, i.e., the same set of samples is utilized to convey information of more than one message bit. This embedding technique enables this scheme to attain higher embedding capacity. During embedding, high audio quality is maintained by controlling distortion level of the sample groups which is later verified by the objective measurement of quality. Security of the proposed system is ensured by the random embedding positions and associated information which cannot be determined statistically without the secret key. Error correction code is also incorporated in this scheme to enhance its robustness against external attacks. At the detector, watermark extraction is performed by informed detection while utilizing correlation coefficient. Eventually, several signal manipulation attacks, according to different benchmarks, are applied to evaluate performance of the proposed scheme. The test results ensure high retrieval rate of the watermark message in most of the instances.

Modification of the loop filter design for a plucked string instrument

A modified loop filter design method for plucked string instruments is introduced. Previous loop filter designs do not properly represent the frequency-dependent damping of a silk stringed instrument. To solve this problem, the modified method evaluates how many harmonics are required and which portion of the sound should be chosen to effectively replicate the instrument. Then, the most reasonable filter parameters are determined based on the frequency signal-to-noise ratio of the synthesized sound.