China 
Africa 
Explore chapters and articles related to this topic
Scalp EEG Classification Using TQWT-Entropy Features for Epileptic Seizure Detection
Published in Mridu Sahu, G. R. Sinha, Brain and Behavior Computing, 2021
Komal Jindal, Rahul Upadhyay, Prabin Kumar Padhy, Hari Shankar Singh
Various time-frequency transforms viz. Stockwell transform (ST) [7], Chirplet transform, and discrete wavelet transform (DWT) are commonly utilized EEG-based time-frequency analysis methods for epilepsy diagnosis. However, many of these transforms have their own limitations [8]. Compared to other time-frequency based methods, many researchers have worked on wavelet based time-frequency decomposition and developed efficient pre-processing and feature extraction methodologies for biomedical applications [9]. Studies carried out on wavelet transform (WT) suggest that the Q-factor value should be low for less oscillatory EEG activity. On the contrary, the Q-factor value should be high for highly oscillatory physiological activity [10]. However, the limitation of WT lies in its incapability to tune Q-factor value as per constraints [11]. Also, the selection of an appropriate mother wavelet function is a challenging task.
Speech feature extraction using linear Chirplet transform and its applications*
Published in Journal of Information and Telecommunication, 2023
Hao Duc Do, Duc Thanh Chau, Son Thai Tran
Figures 2 and 3 show the instantaneous frequency trajectories returned by Linear Chirplet Transform with different chirp rates. Figure 2 presents the trajectory with , corresponding with Fourier Transform. On the other hand, Figure 3 shows the trajectory with a chirp rate . These illustrations mean that Fourier Transform is a special case of LCT with zero chirp rate. Moreover, LCT can transform in many directions with the flexibility of α.
Gear fault diagnosis using an improved Reassigned Smoothed Pseudo Wigner-Ville Distribution
Published in Cogent Engineering, 2018
Dennis Hartono, Dunant Halim, Gethin Wyn Roberts
In addition to time domain-based analyses, the time-frequency analysis has been attracted a lot of attentions in the field of gear fault diagnosis due to its capability in providing a simultaneous representation of time and frequency information. The effective identification of characteristic frequencies of the gearbox, such as gearmesh frequency, its harmonics and sidebands, can provide valuable fault diagnosis information. Although the simultaneous time-frequency information provided by JTFA-based methods has been useful in determining the state-of-health of the gearbox (Baydar & Ball, 2000), a number of limitations still need to be addressed. An example of a widely-used JTFA method for gear fault diagnosis is the Continuous Wavelet Transform (CWT) (Staszewski & Tomlinson, 1994). Its capability to provide multi scale features in a single plot offers an advantage compared to the Short Time Fourier Transform (STFT) (Hartono, Halim, Roberts, & Liu, 2016; Yang, Peng, Meng, & Zhang, 2012). However, it is still restricted by the time duration-frequency bandwidth principle which restricts the simultaneous fine time-frequency localization. Other signal processing methods that can provide an informative time-frequency representation, such as the Chirplet Transform and Warblet Transform, have been recently proposed. However, their current algorithms are primarily restricted for analyzing mono-component signals only (Yang et al., 2012). Moreover, the synchro-squeezing method, which is an improvement of the CWT method, has also been proposed, although it is not designed to analyze transient signals of short temporal duration (Iatsenko, McClintock, & Stefanovska, 2016) that corresponds to the fault characteristics of a faulty gearbox. The other JTFA method, the Wigner-Ville Distribution (WVD) has been proven to have the finest time-frequency localization compared to other existing JTFA method. However, due to its bilinear property, it has the cross term interference (CTI) when analyzing multi-components signal that can mislead the interpretation. Hence, it is not possible to apply such a method to real experimental data, particularly for gearbox signals that do not only contain multi-components signals with its harmonics of gearmesh frequency and sidebands, since the signals are generally corrupted by heavy measurement noise. Therefore, a number of methods have been developed in the last three decades with the attempts to alleviate the CTI of WVD. One of the most versatile methods is the Smoothed Pseudo Wigner-Ville Distribution (SPWVD) method (Baydar & Ball, 2003), although it should be noted that the CTI reduction is done at the expense of the time-frequency localization in WVD (Auger et al., 2013). In this case, a smearing phenomenon in the SPWVD is expected to happen and this is a crucial problem for analyzing multi-components signal like a gearbox signal where several sidebands around the gearmesh frequency can appear at the same time. The SPWVD method has been demonstrated to be effective in removing the CTI of WVD for gear fault diagnosis purposes (Baydar & Ball, 2003). Nevertheless, regardless the benefit of using the existing SPWVD method, a further improvement of WVD method is still highly desirable for achieving more accurate gear fault diagnosis.