All submissions of the EM system will be redirected to Online Manuscript Submission System. Authors are requested to submit articles directly to Online Manuscript Submission System of respective journal.

A New Method in Classification using Stationary Transformed Wavelet Features and Moments

Arash Kalami*

Department of Electrical Engineering, Urmia Branch, Islamic Azad University, Urmia, Iran

*Corresponding Author:
Arash Kalami
Department of Electrical Engineering
Urmia Branch, Islamic Azad University, Urmia, Iran

Received date: 13/05/2014; Revised date: 18/06/2014; Accepted date: 24/06/2014

Visit for more related articles at Research & Reviews: Journal of Pure and Applied Physics


In order to study the feasibility of utilizing wavelet features for signal feature extraction and generation, we perform Method in Classification using Stationary Transformed Wavelet Features and Moments. A signal feature extraction and generation experiment by sequential employment of the wavelet feature based signal on each frame of a video Signal detector based on wavelet feature has problem to treat large Image classification variations. On the successful transformed features case, a rectangle shows the found signal region


Image classification, database, feature extraction and generation, MPEG7, pattern recognition


Effect of the database, when run on the well known testing sequence used in coding community. This sequence contains frames. The detector works well with near frontal signal and handles small rotations, but fails in case the rotation is large. It shows part of the transformed features result. The two outputs, refers to finding a signal and not finding a signal respectively [1-3]. From the whole curve, we estimated that around the frames have not been correctly reported as having a signal. Transformed features rate of the signal transformed features database. It stands for one signal found, it stands no signal found by the database. Altogether frames are reported as having a signal. Most failures are simply due to the lack of training data with signals under large rotations, since the classifiers used in signal transformed features were trained by learning from near frontal signals. One may argue that the database can be learned to cover signals with a large rotation by adding more signals with large rotation into the training data set. This dramatically increase the difficulty of the learning process. In fact, it is the rigid spatial coupling of the individual Wavelet features that limits its Image classification coverage. we show that if we relax the rigid connection between features, that is, if we adopt a deformable signal graph [4,5], we can handle the case of large rotations. Before feature extraction and generation, a set of facial features has to be selected. a total of significant coefficients are used for transformed features, which gives good results. Since our task is to track a specific signal, personal feature response patterns should be adopted to get a reliable feature extraction and generation. The feature extraction and generation consists of finding the best matching deformable signal graph in the new video frame. An example of a deformable signal graph is shown in Through determine the search path for matching, and also transform the problem from Rotation Invariant Moments into pattern. We constrain the deformation in sense of distance changing. Other constraints, such as relative angle is also possible choice.

Methods and Materials

We build a feature pool based on the work done in where we only select significant features captured by the signal transformed features database. An individual feature is selected if its response magnitude is above a predefined threshold. It shows a typical feature set selected in our database. The white rectangles are the wavelet features. One rectangle includes both white and neighboring black rectangles, but the black rectangles are not drawn for visualization Image classification. Selected wavelet facial features set for feature extraction and generation. The Original signal graph example the deformed graph. We only use prominent wavelet features with a big response to the detected signal. A signal graph is formed by spatially arranging all selected wavelet features. The signal graph will later be used for signal feature extraction and generation. Dynamic Programming is used for matching signal graphs extracted from different video frames. A major difference between the new scheme and the old scheme lies in the treatment of the selected feature set. In the old scheme, the features are rigidly connected, while a certain shift between features is allowed in the new scheme. Another difference is that a simple selection of features is used, that is, only the prominent features whose response have big magnitudes are adopted. When the signal transformed features database fails to detect the signal due to big rotation of signal, Dynamic Programming could be used to search for an optimal matching between the new frame and previous frame in which the signal is detected. In this way, the signal area is found through a matching process and not by a transformed features process, and we do not assume any knowledge of the signal in this process. The feature extraction and generation of a wavelet feature set. It is typically used in one- dimensional (pattern) matching problem. In order to use it for Rotation Invariant Moments problem, a straightforward way would be to convert the Rotation Invariant Moments problem into a pattern case. One simple way is to use a matching path to link all the Rotation Invariant Moments components. Although we would lose one dimension of constraints in such an adaption, we still could be able to sustain the Rotation Invariant Moments constraints somewhat if we carefully select the matching path. Figure 1 shows Sample image of MPEG-7 Database. Since the orthographic projection of motion into image plane could be approximated by an affine transform, the deformation of the feature set, in our case, the layout of rectangles thus should move in a “uniformed” way due to this constraint of affine transform. This means the pattern matching path can cross itself when passing through all components. The path should be a shortest path, or at least an economic path in the sense that the accumulated deformation should be confined with the individual deformation among components. A circulated path, for example, could hardly the Different graph structures could serve classification, like a continuous curve path, or a path. Although a spanning tree structure seems more suitable for representing the signal structure, we tried only the simple curve path in this work. An example of signal graph path, which connects center of rectangles in a low-cost order. To find a single path to link all components is exactly the same problem and can adapt the well-known Simulated Annealing method for searching such a path. A linked feature set consists of the deformable signal graph. we again use the Dynamic Programming technique, for the matching Image classification. The trellis structure of the searching path used by the algorithm.


Figure 1: Sample image of MPEG-7 Database

Result and Discussion

The horizontal axis represents a set of ordered pattern recognition. It refers to the wavelet features in the path. The task is to find a global optimal matching between the template image features and new input image features. It allows deformation during search, where It refers to template features and I to the new input features. This could be done via allowing deformation between each pair of feature components. It shows two rectangles in template, and after the deformation. The range of deformation between each pair is depending on their distance, and a pattern recognition sequence is only allowed when their distance satisfy the relation where is the deformation range, and is the distance between the original template feature pair. In our experiments, we choose to allow MPEG-7 database of deformation, that A local match at pattern recognition and location is the distance between the template feature response and input feature response: where is a function of the distance, the bigger the distance, the smaller should be, is also called a local matching cost. The function stand for the response for the wavelet features. In our work, we choose to be an exponential function of the distance: is a constant coefficient for normalization which is set, the curvature. The goal is find the allowed deformation that maximizes the sum of all pattern recognition matches, or a global cost: and the search is performed along the ordered pattern recognition path. The search includes the following steps: At pattern recognition, check the matching costs in previous step. Those costs are saved in all sites in the previous pattern recognition. For each site, determine the search according to allowed deformation range. For each site, take the largest cost within the radius as its own cost step, and save the previous node. This step encodes global the deformation information into the search process. It does the accumulation and advances the accumulated score forward in the trellis. Perform the local matching and evaluate match cost for all possible sites in current pattern recognition. The total matching cost up to the current pattern recognition could be calculated as the accumulated cost in pattern recognition are coefficients for the local match and the global match respectively. These coefficients could be adjusted to control the deformation behavior. Unfortunately there is no known way to find the optimal values. They have to be adjusted through empirical experience. When the search approaches the end pattern recognition, the node with the highest score at end pattern recognition is extracted and the stored list of visited nodes are backtracked to find the matching path. One such path represents a matched deformed signal graph. The algorithm always finds a global, optimal matching. Due to the accumulated deformation effect, at the end, the deformation could be rather big and strange feature pattern could be expected. Figure 2 shows Classified image using Image classification method. One common trick to alleviate this problem is to constrain the distance between the first pattern recognition and last pattern recognition. We set up a database for signal transformed features and feature extraction and generation, based on our deformable signal graph algorithm. one for signal transformed features and one for signal feature extraction and generation. Both the signal detector and the module make use of wavelet features. The pre-computation for the integral image needs to be done only once.


Figure 2: Classified image using Image classification method


In our experiment, both the module and Dynamic Programming module search the whole image area in Classification using Stationary Transformed features and moments. The database runs at an average speed of video frames per second on frame size pixels. In this experiment, we tried to detect the position of the signal region. The aim is to check how well handles Image classification variation. frame Computing Detect Block diagram of our Signal detecting database.


This paper is the result of a research project approved by research committee at Urmia Branch, Islamic Azad University, Urmia, Iran. Authors would like to thank Urmia Branch, Islamic Azad University for all supports.