Wavelet Packet Method Research Articles

An ensemble method of full wavelet packet transform and neural network for short term electrical load forecasting

Due to high penetration of distributed energy resources, integration of intermittent renewable energy resources and deployment of demand-side management, highly accurate short-term load forecasting becomes increasingly important. This paper proposes a full wavelet neural network approach for short-term load forecasting, which is an ensemble method of full wavelet packet transform and neural networks. The full wavelet packet transform model is used to decompose the load profile and various features into several components with different frequencies and these components are used to train the neural networks. To perform load forecasting, the full wavelet packet transform model decomposes features into various components that are fed into the trained neural networks, and the outputs of the neural networks are constructed as the forecasted load. The proposed model is applied for load prediction in the electric market of Ontario, Canada. Simulation results show that the proposed approach reduces the mean absolute percentage error (MAPE) by 20% in comparison with the traditional neural network method. The proposed approach can be used by utilities and system operators to forecast electricity consumption with high accuracy, which is highly demanded for renewable energy integration, demand-side management and power system operation.

Feature Extraction and Classification of Lower Limb Motion Based on sEMG Signals

Surface electromyography (sEMG) signals can reflect the body motion information and are widely used in military, medical rehabilitation, industrial production. The lower limb motion classification mainly includes feature extraction and classification model establishment. Firstly, we proposed a feature extraction method based on the wavelet packet transform (WPT) and principal component analysis (PCA). We used the wavelet packet method to decompose the sEMG signals of three muscles in the lower limb and got the 24-dimensional eigenvector. To reduce the calculation and improve the speed of the classification model, we used the PCA method to reduce the dimension of the feature vector and got the 3-dimensional eigenvector. Then, we proposed a method based on the scale unscented Kalman filter (SUKF) and neural network (NN) for lower limb motion classification. Through the scale correction unscented transform (SCUT) could optimize the neural network weight and improve lower limb motion classification accuracy. Finally, the experimental results showed that the average accuracy was 93.7%. Compared with the backpropagation neural network (BPNN) and wavelet neural network (WNN), this method could improve the accuracy and reliability of the lower limb motion classification.

Extraction of the instantaneous speed fluctuation based on normal time–frequency transform for hydraulic system

The instantaneous speed fluctuation is the most direct manifestation of system dynamics behavior, which reflects the running status of the hydraulic system. It is crucial to mine the information contained in instantaneous speed fluctuation and extract them accurately. This paper first introduces the instantaneous speed measurement of the hydraulic system, and it is found that the instantaneous speed signal with an equal angle sampling is a harmonic signal. Next, causes of instantaneous speed fluctuation for hydraulic system are derived by kinetics analysis, where the working mechanism of the axial piston motor and the inevitable misalignment of the shafts are concerned. Then the normal time–frequency transform and the inaction method are introduced, and its high accuracy for harmonic signal extraction in additive noise is verified by comparing with the wavelet packet method with simulation analysis. Finally, the validity of the cause analysis and the extraction method are verified by experiments under different working conditions. The experimental results show that the frequencies of the instantaneous speed fluctuation components can be displayed automatically by normal time–frequency transform spectrum and their waveforms can be extracted accurately by the inaction method. It is also found that the change of the rotating speed and torque can be presented by the variety of the magnitude of the instantaneous speed fluctuation components at the 7th order and the 14th order, respectively. Cause analysis and accurate extraction of the instantaneous speed fluctuation will provide theoretical support and data support for the performance evaluation, state monitoring, and fault diagnosis of the hydraulic system.

Detection of Tool Wear in Drilling CFRP/TC4 Stacks by Acoustic Emission

Composite stacked material is widely used in many fields, such as aircraft, rocket, missile, and soon. In the process of drilling carbon fiber-reinforced polymer and titanium alloy lamination materials (CFRP/TC4, which is a typical structure in aircraft), tool wear is quick and serious, because the machining condition of the respective layers in stack is different. To ensure the drilling quality, drilling tool needs to be changed frequently. Therefore, if there is a detecting system in automatic drilling device to monitor and predict the tool wear effectively, it will help to improve the production efficiency and reduce the tool cost. Based on the experiments of drilling CFRP/TC4 stacks, the acoustic emission (AE) signals were carefully analyzed using the method of statistical analysis, spectrum analysis, and wavelet packet. The results show that the root-mean-square value of the AE signals and the energy of the wavelet packet are correlated with the tool wear. Meanwhile, experiments indicate that chips and tool fracture will cause instantaneous signal mutation, which may be appeared as disturbances. This has increased the difficulty of identifying the tool wear. With repeated experiments and comparison, it was found that the percentage of wavelet packet energy of some certain frequency sections increased with the tool wear, while some decreased. These variations will be possible to be used as features for further prediction in monitoring system.

Experimental study on micro-damage identification in reinforced concrete beam with wavelet packet and DIC method

Abstract Structural damage identification has become a research hotspot in the field of science and engineering. Previous predominant lines of research on structural damage identification methods focused on macroscopic damage, which is visible to naked eyes, but there are few studies on structural micro-damage identification. Towards this end, this paper proposes a combination of digital image correlation (DIC) technique and wavelet packet analysis for micro-damage identification and to achieve real-time early warning of micro-damage initiation in concrete structures. This work implements a four-point bending test of reinforced concrete beams, as well as adopts the step loading and unloading method, following which, the beam acceleration response signal and the damage strain field are collected. Subsequently, the collected response and excitation are processed to obtain the excitation response function, reflecting the dynamic structural state. The wavelet packet is used to process the excitation response function and image under each load level to obtain the corresponding energy spectrum. In this way, the respective micro-damage identification indices are calculated for comparison. The experimental results show that the proposed methodology can properly characterize the damage state of structures under various stages of stress, provide a warning for the initiation of micro-damage, and even accurately determine the location and degree of micro-damage, based on the damage strain field.

A Bayesian wavelet packet denoising criterion for mechanical signal with non-Gaussian characteristic

Three aspects of the traditional Bayesian wavelet packet denoising method have been rarely discussed in the literature for mechanical signals: (1) how to reduce noise if a precise prior description is needed for non-Gaussian data; (2) how to programmatically select a reasonable decomposition level; and (3) how to evaluate the denoising effect from multiple factors. Such three aspects have restricted the enhancement ability of the Bayesian denoising method. To seek the potential of application, this paper tries to establish a set of criteria. It develops the general Bayesian wavelet packet method based on the minimum mean square error, designs two indexes to select the decomposition level, and initially presents how to judge the denoising effect by multi-factor analysis. This criterion has two advantages: (1) if needed, any prior distribution can be employed to match with the signals; and (2) the selected decomposition level can maximize the distinction between real data and noise data. A comparison test is expected to show this enhancement ability of the proposed criterion. Simulated and experimental data demonstrate its excellent performance. This novel method keeps interpretable time-domain features and shows better denoising performance for weak information recognition than the published methods.

Research on road extraction of remote sensing image based on convolutional neural network

Road is an important kind of basic geographic information. Road information extraction plays an important role in traffic management, urban planning, automatic vehicle navigation, and emergency management. With the development of remote sensing technology, the quality of high-resolution satellite images is improved and more easily obtained, which makes it possible to use remote sensing images to locate roads accurately. Therefore, it is an urgent problem to extract road information from remote sensing images. To solve this problem, a road extraction method based on convolutional neural network is proposed in this paper. Firstly, convolutional neural network is used to classify the high-resolution remote sensing images into two classes, which can distinguish the road from the non-road and extract the road information initially. Secondly, the convolutional neural network is optimized and improved from the training algorithm. Finally, because of the influence of natural scene factors such as house and tree shadow, the non-road noise still exists in the road results extracted by the optimized convolutional neural network method. Therefore, this paper uses wavelet packet method to filter these non-road noises, so as to accurately present the road information in remote sensing images. The simulation results show that the road information of remote sensing image can be preliminarily distinguished by convolutional neural network; the road information can be distinguished effectively by optimizing convolutional neural network; and the wavelet packet method can effectively remove noise interference. Therefore, the proposed road extraction method based on convolutional neural network has good road information extraction effect.

Speech detection enhancement in optical fiber acoustic sensor via adaptive threshold function

Abstract In this paper, we demonstrate a novel speech enhancement method for improving detection performance of optical fiber acoustic sensor. The proposed method is based on a novel threshold function with adaptive threshold for wavelet analysis. For a certain speech signal, the adaptive threshold in the function is set lower in speech presence region and higher in speech absence region, which ensures more speech details preserved as well as noise reduced. At the same time, the new threshold function guarantees the continuity and no deviation of speech signal. Experimental results show that the proposed method is more suitable for optical fiber acoustic sensor to extract speech signal compared with universal wavelet packet method, and it improves speech quality significantly. The clear speech can be obtained via the optical fiber acoustic sensor even if work surroundings of sensing unit is noisy.

Gear fault diagnosis method based on wavelet-packet independent component analysis and support vector machine with kernel function fusion

Aiming at the problem of gear fault diagnosis, in order to effectively extract the features and improve the accuracy of gear fault diagnosis, the method based on wavelet-packet independent component analysis and support vector machine with kernel function fusion is proposed in this research. The proposed wavelet-packet independent component analysis feature extraction method can effectively combine the advantages of wavelet packet and independent component analysis methods and acquire more comprehensive feature information. Besides, the proposed kernel-function-fusion support vector machine can well integrate the advantage characteristics of each kernel function. The energy features of wavelet packet coefficients are acquired with four-layer wavelet packet decomposition and then the extracted energy features are further optimized by the independent component analysis method. The kernel-function-fusion support vector machine method is adopted to realize the gear fault diagnosis. Two kernel function models with the best self-classification accuracy are employed to serve the gear fault diagnosis corporately. The test samples are primarily classified by the main kernel function model, and then some samples are selected to be reclassified with the other kernel function model. Finally, the two kernel function models cooperate to determine the type of test samples. The comparison investigations demonstrate that the proposed method based on wavelet-packet independent component analysis and support vector machine with kernel function fusion achieves very high diagnosis accuracy.

Analysis of the dynamic characteristics of a suspension bridge based on RTK-GNSS measurement combining EEMD and a wavelet packet technique

To investigate the dynamic characteristics of a self-anchored suspension bridge (i.e. Tianjin Fumin Bridge), a real time kinematic-global navigation satellite system (RTK-GNSS) is used to achieve the displacement responses of the structure. Seeing the defect in the positioning accuracy of RTK-GNSS, a combined method (EEMDWP) of ensemble empirical mode decomposition (EEMD) and wavelet packet (WP) analysis is firstly put forward to improve the precision of the vibration signals. Subsequently, fast Fourier transform and the random decrement technique are applied to estimate the natural frequency and the corresponding damping ratio of the structure. Meanwhile, to contrast the field measurement results, the finite element model (FEM) of the structure is established. Finally, the analysis results indicate that: (1) the RTK-GNSS technique is a powerful tool for monitoring the deformation of long-span bridges; (2) the proposed EEMDWP method is demonstrated to be better than the single EEMD or WP method; (3) the structural dynamic parameters are successfully obtained (i.e. the first natural frequency: 0.5873 Hz, the corresponding damping ratio: 2.12%); (4) the results from the field measurement are in agreement with the FEM with a difference of about 5% from each other.

Identifying Degenerative Brain Disease Using Rough Set Classifier Based on Wavelet Packet Method

Population aging has become a worldwide phenomenon, which causes many serious problems. The medical issues related to degenerative brain disease have gradually become a concern. Magnetic Resonance Imaging is one of the most advanced methods for medical imaging and is especially suitable for brain scans. From the literature, although the automatic segmentation method is less laborious and time-consuming, it is restricted in several specific types of images. In addition, hybrid techniques segmentation improves the shortcomings of the single segmentation method. Therefore, this study proposed a hybrid segmentation combined with rough set classifier and wavelet packet method to identify degenerative brain disease. The proposed method is a three-stage image process method to enhance accuracy of brain disease classification. In the first stage, this study used the proposed hybrid segmentation algorithms to segment the brain ROI (region of interest). In the second stage, wavelet packet was used to conduct the image decomposition and calculate the feature values. In the final stage, the rough set classifier was utilized to identify the degenerative brain disease. In verification and comparison, two experiments were employed to verify the effectiveness of the proposed method and compare with the TV-seg (total variation segmentation) algorithm, Discrete Cosine Transform, and the listing classifiers. Overall, the results indicated that the proposed method outperforms the listing methods.

Recovery of facial expressions using functional electrical stimulation after full-face transplantation

BackgroundWe assessed the recovery of 2 face transplantation patients with measures of complexity during neuromuscular rehabilitation. Cognitive rehabilitation methods and functional electrical stimulation were used to improve facial emotional expressions of full-face transplantation patients for 5 months. Rehabilitation and analyses were conducted at approximately 3 years after full facial transplantation in the patient group. We report complexity analysis of surface electromyography signals of these two patients in comparison to the results of 10 healthy individuals.MethodsFacial surface electromyography data were collected during 6 basic emotional expressions and 4 primary facial movements from 2 full-face transplantation patients and 10 healthy individuals to determine a strategy of functional electrical stimulation and understand the mechanisms of rehabilitation. A new personalized rehabilitation technique was developed using the wavelet packet method. Rehabilitation sessions were applied twice a month for 5 months. Subsequently, motor and functional progress was assessed by comparing the fuzzy entropy of surface electromyography data against the results obtained from patients before rehabilitation and the mean results obtained from 10 healthy subjects.ResultsAt the end of personalized rehabilitation, the patient group showed improvements in their facial symmetry and their ability to perform basic facial expressions and primary facial movements. Similarity in the pattern of fuzzy entropy for facial expressions between the patient group and healthy individuals increased. Synkinesis was detected during primary facial movements in the patient group, and one patient showed synkinesis during the happiness expression. Synkinesis in the lower face region of one of the patients was eliminated for the lid tightening movement.ConclusionsThe recovery of emotional expressions after personalized rehabilitation was satisfactory to the patients. The assessment with complexity analysis of sEMG data can be used for developing new neurorehabilitation techniques and detecting synkinesis after full-face transplantation.

A Pilot Study Based on Cerebral Hemoglobin Information to Classify the Desired Walking Speed

To achieve more intelligent performance for walking-assistive devices, spontaneous motion intention of walking speed should be identified for providing a control command. In this letter, cerebral hemoglobin information was analyzed to recognize three levels of walking speed: low, medium, and high speed. Thirty subjects performed walking tasks in three levels of speed, with cerebral hemoglobin information being recorded and oxygenated hemoglobin (oxyHb) mainly analyzed. A novel method of time–frequency–space analysis is proposed to extract features through two steps. First, a wavelet packet method is used to analyze the oxyHb signals in four frequency bands [(I) 0–0.03 Hz, (II) 0.03–0.06 Hz, (III) 0.06–0.09 Hz, and (IV) 0.09–0.12 Hz]. Second, spatial features were summarized for each frequency band. The support vector machine method is used to identify speed levels. The dataset of 15 people is used to train the model, and the dataset of the other 15 people is used for validation. The final recognition rates of low, medium, and high speed are 93.33%, 66.67%, and 80%, respectively. The average recognition is up to 80%. The results indicate that the proposed method of time–frequency–space analysis is feasible for recognizing expected walking speed, and cerebral hemoglobin information could reflect humans’ spontaneous motion intention. Moreover, it may provide a more intelligent control method for walking-assistive devices and promote its development in the future.

Analysis of AE characteristics of tool wear in drilling CFRP/Ti stacked material

Based on the signal data acquired in drilling process of carbon fiber reinforced polymer/titanium alloy (CFRP/Ti) stacked materials, the acoustic emission (AE) characteristic values were carefully studied, by using the method of statistical analysis, spectrum analysis and wavelet packet. The results show that the root mean square(RMS) value of the AE signals and the energy of the wavelet packet are closely related to the tool wear. Meanwhile, experiments indicate that different materials, chips and tool tipping will cause instantaneous signal mutation, which has different forms in time domain and in time-frequency domain. These mutations may increase the difficulty of identifying the tool wear. Fortunately, with repeated experiments and comparison, some identifiable mutations were recognized. When a tool is processed from CFRP to Ti, the signal intensity decreases generally, the high-frequency component of signal increases gradually, and the signal has a tendency to show in high frequencies.

Prediction of bearing damage in wind turbines based on the quadratic root mean square of sub-band manifold

When a high-power wind turbine runs in normal status, and if bearing current exists for a long time, multiple point corrosion would occur and gradually increase, eventually forming a ripple groove on the inner ring race, outer ring race and rolling body. It would lead to more vibration and shock, thereby causing fault in the equipment; the best way to prevent the this kind of fault is to find the effective fault characteristics and predict the damage’s degree on the bearing. In this paper, an adaptive neural network prediction method based on the quadratic root mean square of sub-band manifold is proposed. The damage characteristics can be analyzed by following steps: firstly, the vibration signal is decomposed into multidimensional time frequency space by wavelet packet method. Secondly, the sub-band of the manifold is constructed. The third step is to extract the root mean square value. Finally, the damage characteristics of the bearing current of the two square root sub-band manifold are obtained. Based on the back propagation network, the adaptive prediction model is built, and the training speed could be adjusted automatically according to the prediction error and precision. According to the bearing’s fault mechanism with current damage on the high-power wind turbine, one fault experiment platform has been built to simulate the current damage process of the bearing and verify the prediction method based on the quadratic root mean square of sub-band manifold. The experimental results show that the method can effectively predict the degree of bearing current damage, and the relative error of prediction is less than 5%.

An adaptive denoising fault feature extraction method based on ensemble empirical mode decomposition and the correlation coefficient

Vibration signal processing is commonly used in the mechanical fault diagnosis. It contains abundant working status information. The vibration signal has some features such as non-linear and non-stationary. It has a lot of interference information. Fault information is vulnerable to the impact of the interference information. Empirical mode decomposition denoising method and kurtosis correlation threshold have been widely used in the field of fault diagnosis. But the method mainly depends on the subjective experience, the large number of attempts, and lack of adaptability. In this article, the signals are decomposed into several intrinsic mode functions adaptively with ensemble empirical mode decomposition. The intrinsic mode functions containing the main fault information are selected by the correlation coefficient to emphasize the fault feature and inhibit the normal information. Finally, the energy features of these intrinsic mode functions are taken as inputs of a neural network to identify the fault patterns of rolling bearing. The experiment shows that the neural network diagnosis method based on ensemble empirical mode decomposition has a higher fault recognition rate than based on empirical mode decomposition or wavelet packet method.

Assessment of Emotional Expressions after Full-Face Transplantation.

We assessed clinical features as well as sensory and motor recoveries in 3 full-face transplantation patients. A frequency analysis was performed on facial surface electromyography data collected during 6 basic emotional expressions and 4 primary facial movements. Motor progress was assessed using the wavelet packet method by comparison against the mean results obtained from 10 healthy subjects. Analyses were conducted on 1 patient at approximately 1 year after face transplantation and at 2 years after transplantation in the remaining 2 patients. Motor recovery was observed following sensory recovery in all 3 patients; however, the 3 cases had different backgrounds and exhibited different degrees and rates of sensory and motor improvements after transplant. Wavelet packet energy was detected in all patients during emotional expressions and primary movements; however, there were fewer active channels during expressions in transplant patients compared to healthy individuals, and patterns of wavelet packet energy were different for each patient. Finally, high-frequency components were typically detected in patients during emotional expressions, but fewer channels demonstrated these high-frequency components in patients compared to healthy individuals. Our data suggest that the posttransplantation recovery of emotional facial expression requires neural plasticity.

Energy spectrum analysis of blast waves based on an improved Hilbert–Huang transform

Using the improved Hilbert–Huang transform (HHT), this paper investigates the problems of analysis and interpretation of the energy spectrum of a blast wave. It has been previously established that the energy spectrum is an effective feature by which to characterize a blast wave. In fact, the higher the energy spectra in a frequency band of a blast wave, the greater the damage to a target in the same frequency band. However, most current research focuses on analyzing wave signals in the time domain or frequency domain rather than considering the energy spectrum. We propose here an improved HHT method combined with a wavelet packet to extract the energy spectrum feature of a blast wave. When applying the HHT, the signal is first roughly decomposed into a series of intrinsic mode functions (IMFs) by empirical mode decomposition. The wavelet packet method is then performed on each IMF to eliminate noise on the energy spectrum. Second, a coefficient is introduced to remove unrelated IMFs. The energy of each instantaneous frequency can be derived through the Hilbert transform. The energy spectrum can then be obtained by adding up all the components after the wavelet packet filters and screens them through a coefficient to obtain the effective IMFs. The effectiveness of the proposed method is demonstrated by 12 groups of experimental data, and an energy attenuation model is established based on the experimental data. The improved HHT is a precise method for blast wave signal analysis. For other shock wave signals from blasting experiments, an energy frequency time distribution and energy spectrum can also be obtained through this method, allowing for more practical applications.

Noise Suppression in 94 GHz Radar-Detected Speech Based on Perceptual Wavelet Packet

A millimeter wave (MMW) radar sensor is employed in our laboratory to detect human speech because it provides a new non-contact speech acquisition method that is suitable for various applications. However, the speech detected by the radar sensor is often degraded by combined noise. This paper proposes a new perceptual wavelet packet method that is able to enhance the speech acquired using a 94 GHz MMW radar system by suppressing the noise. The process is as follows. First, the radar speech signal is decomposed using a perceptual wavelet packet. Then, an adaptive wavelet threshold and new modified thresholding function are employed to remove the noise from the detected speech. The results obtained from the speech spectrograms, listening tests and objective evaluation show that the new method significantly improves the performance of the detected speech.

Extraction of weak crack signals based on sparse code shrinkage combined with wavelet packet filtering

Abstract Early crack signals in critical infrastructure components of major equipment are hardly to be extracted due to its low signal noise ratio (SNR). A de-noising method combined wavelet packet (WP) technology with sparse code shrinkage (SCS) is proposed in this study. Firstly, WP reconstruction technology is used to reserve the crack signal with a specified frequency range. That is, the signal is decomposed by Meyer wavelet into five layers, and the signal with the frequency range from 187.5 kHz to 609.375 kHz is reserved. Then SCS method removes noise within the specified frequency range. Namely, the probability density function (PDF) of the signal independent coefficients is estimated via the generalized Gaussian model (GGM) in the independent component analysis (ICA) space. The nonlinear de-noising is finished by utilizing maximum a posteriori (MAP) estimate. The results obtained by the combined method are compared with those generated by the SCS method and the WP de-noising method. It demonstrates that the combined method is the best one among the three methods in extracting weak signals. Its output SNR is −2.38 dB and the correlation coefficient (CC) is 0.54 when the input SNR is −20 dB. They are higher than those obtained by the SCS method (SNR −4.46 dB and CC 0.51). The WP method is the worst (SNR −3.54 dB and CC −0.003). Therefore, the combined method is quite suitable for weak signal extraction.