Sangjin Cho

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.

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.

Time-domain audio watermarking using multiple marking spaces

In this paper, a time-domain audio watermarking scheme is proposed where embedding is done in two different marking spaces which are obtained from the host audio by exploiting the properties of Polar coordinate system. This technique has the advantage of higher embedding capacity due to its double utilization of the same set of audio samples during insertion of watermark message. Simulation results confirm its robustness against attacks like compression, filtering, re-quantization, re-sampling, normalization and white noise addition.

A background subtraction method using color information in the frame averaging process

Accurate and reliable detection of moving object from a video sequence is often hard to do in real case. The most common approach is image segmentation which identifies the object from the video sequence that differs significantly from a background model. Among the other segmentation methods background subtraction is the effective and efficient one. A background subtraction method must gratify some challenges, first of all the method must diminish the noise from the sensor that results alteration in pixel color in the probable background from the original in the current video sequence. Secondly the method must automatically fiddle with the changing scene conditions. All these challenges must satisfy in real time with lower computational complexity. In this paper we represent a background subtraction technique that learns the variation of each pixel throughout the video in terms of average and average differences of pixels color of current frame with the previous one. The BGR (Blue, Green and Red) image is first converted to HSV (Hue, Saturation and Value) color space. And then the algorithm is applied to the H plane only as H plane is invariant to illumination changes. Morphological operations are done for removing image noises. In the last stage edges are detected to find out whether there is any false detection or not. Our experimental result shows that this method is very much reliable in changing scene conditions and noise elimination.

Fault diagnosis system using LPC coefficients and neural network

As rotating machines perform an important role in industrial applications, many researchers have developed various condition monitoring system and fault diagnosis system by applying various techniques such as signal processing and pattern recognition. Recently, fault diagnosis systems using artificial neural network have been proposed. This paper proposes the neural-network-based fault diagnosis system using the proper feature vectors by LPC (linear predictive coding) coefficients. This method has not been reported yet. For the effective fault diagnosis, a MLP (multi-layer perceptron) network is used. From the experiment results, the proposed system shows a perfect fault diagnosis for each faulty case.

Maximizing Incipient Fault Signatures of Rotating Machines Using WE and CLES

ABSTRACT This paper presents a constant speed rotating machines fault analysis method to maximize incipient fault signatures based on computed order tracking with adaptive filter and cyclic logarithmic envelope spectrum (CLES) tools. First, adaptive filter reduces not-concerned harmonic components and measurement artefacts from a vibration signal. To accurately and precisely select the step size in the process of the LMS-type adaptive filter, a new method is proposed based on wavelet entropy (WE). Then, a tacho signal is generated from vibration signal and computed order tracking is performed using this tacho signal. Finally, to reduce the effect of second-order cyclostationarity, CLES tool is used to analyze the incipient fault features. Simulations are carried out and the performances of WE and CLES are compared with traditional method. Furthermore, vibration signals of real experimental data are used to verify the robustness of the proposed method. The experimental results show that the proposed method can effectively maximize the incipient fault signatures of rotating machines.AbstractThis paper presents a constant speed rotating machines fault analysis method to maximize incipient fault signatures based on computed order tracking with adaptive filter and cyclic logarithmic envelope spectrum (CLES) tools. First, adaptive filter reduces not-concerned harmonic components and measurement artefacts from a vibration signal. To accurately and precisely select the step size in the process of the LMS-type adaptive filter, a new method is proposed based on wavelet entropy (WE). Then, a tacho signal is generated from vibration signal and computed order tracking is performed using this tacho signal. Finally, to reduce the effect of second-order cyclostationarity, CLES tool is used to analyze the incipient fault features. Simulations are carried out and the performances of WE and CLES are compared with traditional method. Furthermore, vibration signals of real experimental data are used to verify the robustness of the proposed method. The experimental results show that the proposed method ...

Improved Plucked String Synthesis Model Considering Tuning and Playing Style

ABSTRACT This paper describes a plucked string synthesis model including a simplified tuning system and a right-hand playing style. The proposed model based on digital waveguide synthesis is applied to the Gayageum, a traditional Korean plucked string instrument. The Gayageum has 12 twisted silk strings, each supported by a movable bridge called an Anjok, which is placed on the body and is moved sideways along a string to tune it. The simplified tuning system is formulated with parameters related to the tension of the string and location of the Anjok, and these determine the pitch of a specific string. The unique right-hand playing style, a consecutive flick in which a player flicks a string twice in succession by the middle and index fingers, is implemented by a simple feed-forward filter. The proposed synthesis model is evaluated based on the frequency signal-to-noise ratio.

Audio watermarking alogrithm using subband energy

This paper proposes a novel watermarking algorithm that protects digital copyrights and is robust to attacks. The Algorithm contains two features: finding target subband based on energy and wetting adaptive quantization step-size. These features improve the robustness and preserve high audio quality. Watermarks are embedded in the subband including the significant part of the signal such as a pitch. Generally, the subband containing the pitch has the biggest energy. In order to find this subband, wavelet packet transform is used to decompose the signal into subbands and their energy is calculated. The signal of the selected subbands is transformed in frequency domain using FFT. The watermarks are embedded using quantization index modulation (QIM) for higher values of samples than a certain threshold. Our decoder uses the blind detection using the Euclidean distance between the received signals and the signals using QIM for the received signals. The proposed method shows less than 5% bit error ratio (BER) in the audio watermark benchmarking.

A novel musical synthesizer embedded with physical modeling sound synthesis algorithm

This paper describes a novel synthesizer that can generate sounds of two plucked string and one woodwind instrument. This consists of laser strings, frets, physical modeling sound engine. Laser strings constructed by pairs of a laser module and a photodiode describe plucking, strumming and blowing. Frets are implemented by using voltage divider and they enable the player to represent various playing styles such as guitar solos. Each fret has different voltage values to adjust intonation on the synthesizer. These values are converted into 12-bit data and served as an argument of the sound engine based on digital waveguide model with time-varying propagation speed (DWTVPS). Sound synthesis algorithm is embedded onto the digital signal processor. Consequently, the synthesized sounds are played through a speaker in real-time.

Bearing fault detection and identification using adaptive filter and computed order tracking

Rolling element bearings are the most common component in the domestic and industrial application. Small defects in bearing unless detected in time may cause to malfunctioning of the machinery and also may lead economic loss and even human casualty. This paper proposes a novel method for identification and detection of rolling element bearing faults based on adaptive filter and computed order tracking. Noise and other not-concerned harmonics are reduced from vibration signal using an adaptive filter. And then, computed order tracking is applied to remove the frequency variation of the vibration signal. Finally, spectrum analysis is applied to detect the fault features. Simulation and experimental results indicate that proposed method can identify and detect bearing faults accurately and effectively.