Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120649
Hassaan Haider, J. Shah, Shahid Ikram, Idris Abd Latif
The computationally intensive part of compressed sensing (CS) deals with the sparse signal reconstruction from lesser number of random projections. Finding sparse solution to such an underdetermined system is highly ill-conditioned and therefore requires additional regularization constraints. This research paper introduces a new approach for recovering a K-sparse signal from compressed samples using particle swarm optimization (PSO) along with separable surrogate functionals (SSF) algorithm. The suggested hybrid mechanism applied with appropriate regularization constraints speeds up the convergence of PSO. The estimated original sparse signal is also recovered with great precision. Simulation results show that the signal estimated with PSO-SSF combination outperforms the signal recovery through PSO, SSF and parallel coordinate descent (PCD) methods in terms of reconstruction accuracy. Finally, the efficiency of the proposed algorithm is validated experimentally by exactly recovering a one-dimensional K-sparse signal from only a few number of non-adaptive random measurements.
{"title":"Sparse signal recovery from compressed measurements using hybrid particle swarm optimization","authors":"Hassaan Haider, J. Shah, Shahid Ikram, Idris Abd Latif","doi":"10.1109/ICSIPA.2017.8120649","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120649","url":null,"abstract":"The computationally intensive part of compressed sensing (CS) deals with the sparse signal reconstruction from lesser number of random projections. Finding sparse solution to such an underdetermined system is highly ill-conditioned and therefore requires additional regularization constraints. This research paper introduces a new approach for recovering a K-sparse signal from compressed samples using particle swarm optimization (PSO) along with separable surrogate functionals (SSF) algorithm. The suggested hybrid mechanism applied with appropriate regularization constraints speeds up the convergence of PSO. The estimated original sparse signal is also recovered with great precision. Simulation results show that the signal estimated with PSO-SSF combination outperforms the signal recovery through PSO, SSF and parallel coordinate descent (PCD) methods in terms of reconstruction accuracy. Finally, the efficiency of the proposed algorithm is validated experimentally by exactly recovering a one-dimensional K-sparse signal from only a few number of non-adaptive random measurements.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116313681","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120570
S. T. Lim, M. K. Ahmed, Sungbin Lim
Diabetic macular edema is the leading cause of visual loss for patients with diabetic retinopathy, a complication of diabetes. Early screening and treatment has been shown to prevent blindness in diabetic retinopathy and diabetic macular edema. The Early Treatment Diabetic Retinopathy Study (ETDRS) and the Diabetic Macular Edema Disease Severity Scale are the common screening standards based on the distance of exudates from the fovea. Instead of focusing on the macula region, this research adopts a global approach using texture classification to grade the fundus images into three stages: normal, moderate diabetic macular edema and severe diabetic macular edema. The proposed algorithm starts with a modified completed Local Binary Pattern (CLBP) to extract the image local gray level for all RGB channels. The obtained feature vector will then be fed into a multiclass Support Vector Machine (SVM) for classification. The 100 fundus images selected to be utilized for training and testing set were taken from MESSIDOR and these images were reviewed by an ophthalmologist for cross-validation. The algorithm using the CLBP demonstrates a sensitivity of 67% with a specificity of 30% while the proposed modified CLBP yields a higher sensitivity and specificity of 80% and 70% respectively.
{"title":"Automatic classification of diabetic macular edema using a modified completed Local Binary Pattern (CLBP)","authors":"S. T. Lim, M. K. Ahmed, Sungbin Lim","doi":"10.1109/ICSIPA.2017.8120570","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120570","url":null,"abstract":"Diabetic macular edema is the leading cause of visual loss for patients with diabetic retinopathy, a complication of diabetes. Early screening and treatment has been shown to prevent blindness in diabetic retinopathy and diabetic macular edema. The Early Treatment Diabetic Retinopathy Study (ETDRS) and the Diabetic Macular Edema Disease Severity Scale are the common screening standards based on the distance of exudates from the fovea. Instead of focusing on the macula region, this research adopts a global approach using texture classification to grade the fundus images into three stages: normal, moderate diabetic macular edema and severe diabetic macular edema. The proposed algorithm starts with a modified completed Local Binary Pattern (CLBP) to extract the image local gray level for all RGB channels. The obtained feature vector will then be fed into a multiclass Support Vector Machine (SVM) for classification. The 100 fundus images selected to be utilized for training and testing set were taken from MESSIDOR and these images were reviewed by an ophthalmologist for cross-validation. The algorithm using the CLBP demonstrates a sensitivity of 67% with a specificity of 30% while the proposed modified CLBP yields a higher sensitivity and specificity of 80% and 70% respectively.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"331 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116358646","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120628
R. Zafar, A. Malik, Aliyu Nuhu Shuaibu, M. J. U. Rehman, S. Dass
Experiment design has a key role in the functional magnetic resonance imaging (fMRI) data analyses. Block designs are suitable to localize functional areas but are not able to measure the transient changes in the brain activity. Event related design is a better approach and saves time and resources like single trial analyses. In this study, we explored the event related design with single, and multi trials with different order. In multi trials, instead of using lot of trials, we did analyses with two trials per image. The result suggest that the combination of multiple trials, order of trials and selection of significant voxels can give better results in terms of classification accuracy. Moreover, single and two trials per image saves resources as compared to many trials.
{"title":"Multiple trials in event related fMRI for different conditions","authors":"R. Zafar, A. Malik, Aliyu Nuhu Shuaibu, M. J. U. Rehman, S. Dass","doi":"10.1109/ICSIPA.2017.8120628","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120628","url":null,"abstract":"Experiment design has a key role in the functional magnetic resonance imaging (fMRI) data analyses. Block designs are suitable to localize functional areas but are not able to measure the transient changes in the brain activity. Event related design is a better approach and saves time and resources like single trial analyses. In this study, we explored the event related design with single, and multi trials with different order. In multi trials, instead of using lot of trials, we did analyses with two trials per image. The result suggest that the combination of multiple trials, order of trials and selection of significant voxels can give better results in terms of classification accuracy. Moreover, single and two trials per image saves resources as compared to many trials.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115488130","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120655
Chun Yeen Ho, C. Leow
The existing work on cooperative non-orthogonal multiple access (NOMA) considers one-way relaying which consumes extra channel resources for the relaying operation. The use of extra channel resource results in degradation in spectral efficiency. This paper proposes two-way relaying in cooperative NOMA to enhance the spectral efficiency. The proposed scheme enable two-way information exchange between base station and users without consuming extra channel resource. In additional, the NOMA power allocation region is proposed to achieve a better rate compared to orthogonal multiple access scheme. Based on Monte Carlo simulation, the proposed scheme is shown to achieve better sum rate compared to OMA scheme and conventional cooperative NOMA.
{"title":"Cooperative non-orthogonal multiple access using two-way relay","authors":"Chun Yeen Ho, C. Leow","doi":"10.1109/ICSIPA.2017.8120655","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120655","url":null,"abstract":"The existing work on cooperative non-orthogonal multiple access (NOMA) considers one-way relaying which consumes extra channel resources for the relaying operation. The use of extra channel resource results in degradation in spectral efficiency. This paper proposes two-way relaying in cooperative NOMA to enhance the spectral efficiency. The proposed scheme enable two-way information exchange between base station and users without consuming extra channel resource. In additional, the NOMA power allocation region is proposed to achieve a better rate compared to orthogonal multiple access scheme. Based on Monte Carlo simulation, the proposed scheme is shown to achieve better sum rate compared to OMA scheme and conventional cooperative NOMA.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127288535","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120585
Mohammad F. Jamaluddin, M. F. A. Fauzi, F. S. Abas
Histopathological analysis of tissues has been gaining a lot of interests recently, from developing computer algorithms to assist pathologists from cell detection and counting, to tissue classification and cancer grading. With the advent of whole slide imaging, the field of digital pathology has gained enormous popularity, and is currently regarded as one of the most promising avenues of diagnostic medicine. Deep learning advancement on image set today has successfully evolved as many models has been proposed and produced state-of-the-art object classifying results. This is not limited to large database such as Imagenet but also has seen applications in other medical image analysis related areas. In this paper we have carefully constructed and expanded the deep model network to classify normal and tumor slides in histology images of lymph nodes tissue. We have proposed our own deep learning model based on convolutional neural network with smaller requirement using 64×64×3 input image with 12 convolutional layer with max pooling and ReLU activation function. Our method has better AUC result at 0.94 than the winner of Camelyon16 Challenge with AUC of 0.925.
{"title":"Tumor detection and whole slide classification of H&E lymph node images using convolutional neural network","authors":"Mohammad F. Jamaluddin, M. F. A. Fauzi, F. S. Abas","doi":"10.1109/ICSIPA.2017.8120585","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120585","url":null,"abstract":"Histopathological analysis of tissues has been gaining a lot of interests recently, from developing computer algorithms to assist pathologists from cell detection and counting, to tissue classification and cancer grading. With the advent of whole slide imaging, the field of digital pathology has gained enormous popularity, and is currently regarded as one of the most promising avenues of diagnostic medicine. Deep learning advancement on image set today has successfully evolved as many models has been proposed and produced state-of-the-art object classifying results. This is not limited to large database such as Imagenet but also has seen applications in other medical image analysis related areas. In this paper we have carefully constructed and expanded the deep model network to classify normal and tumor slides in histology images of lymph nodes tissue. We have proposed our own deep learning model based on convolutional neural network with smaller requirement using 64×64×3 input image with 12 convolutional layer with max pooling and ReLU activation function. Our method has better AUC result at 0.94 than the winner of Camelyon16 Challenge with AUC of 0.925.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"130 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121904647","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120654
Jie He, Da-Zheng Feng, Xiao-Jun Yang
In this paper, a hybrid space time adaptive processing (STAP) algorithm of direct data domain (DDD) approach and cost function reconstruction is presented to provide a solution to sample support problem at a low cost of space-time aperture loss. The correlation matrix estimated in DDD approach is partitioned into sub-matrices and two equivalent cost functions are reconstructed. By iteratively solving cost functions, sample support requirements and computational burden can be mitigated. The experiments results on the real data show that the proposed algorithm outperforms conventional DDD method and DDD-JDL with low aperture loss.
{"title":"A modified direct data domain STAP approach based on cost function reconstruction","authors":"Jie He, Da-Zheng Feng, Xiao-Jun Yang","doi":"10.1109/ICSIPA.2017.8120654","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120654","url":null,"abstract":"In this paper, a hybrid space time adaptive processing (STAP) algorithm of direct data domain (DDD) approach and cost function reconstruction is presented to provide a solution to sample support problem at a low cost of space-time aperture loss. The correlation matrix estimated in DDD approach is partitioned into sub-matrices and two equivalent cost functions are reconstructed. By iteratively solving cost functions, sample support requirements and computational burden can be mitigated. The experiments results on the real data show that the proposed algorithm outperforms conventional DDD method and DDD-JDL with low aperture loss.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"2016 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121490559","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120645
M. N. Mohd, Yoosuf Nizam, S. Suhaila, M. M. Jamil
Systems developed to classify human activities to identify unintentional falls are highly demanding and play an important role in our daily life. Human falls are the main obstacle for elderly people to live independently and it is also a major health concern due to aging population. Different approaches are used to develop human fall detection systems for elderly and people with special needs. The three basic approaches used include some sort of wearable devices, ambient based devices or non-invasive vision based devices using live cameras. Most of such systems are either based on wearable or ambient sensor which is very often rejected by users due to the high false alarm and difficulties in carrying them during their daily life activities. This paper proposes a fall detection system based on an algorithm using combination of machine learning and human activity measurements such as changes of human height and rate of change of the subject during any of the activity. Classification of human fall from other activities of daily life is accomplished using height, changes in velocity and acceleration of the subject extracted from the depth information. Finally position of the subject and SVM classification is used for fall confirmation. From the experimental results, the proposed system was able to achieve an average accuracy of 97.39% with sensitivity of 100% and specificity of 96.61%.
{"title":"An optimized low computational algorithm for human fall detection from depth images based on Support Vector Machine classification","authors":"M. N. Mohd, Yoosuf Nizam, S. Suhaila, M. M. Jamil","doi":"10.1109/ICSIPA.2017.8120645","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120645","url":null,"abstract":"Systems developed to classify human activities to identify unintentional falls are highly demanding and play an important role in our daily life. Human falls are the main obstacle for elderly people to live independently and it is also a major health concern due to aging population. Different approaches are used to develop human fall detection systems for elderly and people with special needs. The three basic approaches used include some sort of wearable devices, ambient based devices or non-invasive vision based devices using live cameras. Most of such systems are either based on wearable or ambient sensor which is very often rejected by users due to the high false alarm and difficulties in carrying them during their daily life activities. This paper proposes a fall detection system based on an algorithm using combination of machine learning and human activity measurements such as changes of human height and rate of change of the subject during any of the activity. Classification of human fall from other activities of daily life is accomplished using height, changes in velocity and acceleration of the subject extracted from the depth information. Finally position of the subject and SVM classification is used for fall confirmation. From the experimental results, the proposed system was able to achieve an average accuracy of 97.39% with sensitivity of 100% and specificity of 96.61%.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"4 4","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120978695","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120592
Omar Salem Baans, A. B. Jambek
Microarray analysis is one of the most suitable tools available for scientists concerned with DNA sequences to study and examine gene expression. Through microarray analysis, the gene expression sequence can be obtained and biological information on many diseases can be acquired. The gene expression information contained in the microarray can be extracted using image-processing techniques. Microarray image processing consists of three main steps: gridding, segmentation and intensity extraction. This paper analyses the computational time for this microarray image processing. The results show that the intensity extraction consumes majority of the overall computational time. More detail analysis reveals that this high computational time is due to the background correction part of the process, as discussed in the second part of this paper.
{"title":"Software profiling analysis for DNA microarray image processing algorithm","authors":"Omar Salem Baans, A. B. Jambek","doi":"10.1109/ICSIPA.2017.8120592","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120592","url":null,"abstract":"Microarray analysis is one of the most suitable tools available for scientists concerned with DNA sequences to study and examine gene expression. Through microarray analysis, the gene expression sequence can be obtained and biological information on many diseases can be acquired. The gene expression information contained in the microarray can be extracted using image-processing techniques. Microarray image processing consists of three main steps: gridding, segmentation and intensity extraction. This paper analyses the computational time for this microarray image processing. The results show that the intensity extraction consumes majority of the overall computational time. More detail analysis reveals that this high computational time is due to the background correction part of the process, as discussed in the second part of this paper.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132931602","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120582
E. Jaja, A. Rahman, Z. Omar, M. Zabidi, U. U. Sheikh
This paper presents two intra prediction algorithms in High Efficiency Video Coding (HEVC) encoder for reducing the computational complexity and increase the encoding speed. The first algorithm takes advantage of the high spatial correlation among neighboring pixels to substitute the reference samples as the first row or first and second rows of the current block to be predicted, while the pixels intensities in the remaining rows or columns as in the case of horizontal predictions, are extrapolated as usual. Secondly, due to spatial correlations in video block data among adjacent blocks, it has been established that the intra prediction mode of the current block has a high probability of being a member of the most probable mode set or a slight variation of one of the most probable modes. These algorithms are combined and implemented on the HM16 reference software, and show speedup of 23.4% and 22.7% in encoding time using the all-intra-main configuration, with minimal reduction in bitrate of 0.21% and 0.22% respectively.
{"title":"Intra block prediction using first-second row non-directional samples in HEVC video coding","authors":"E. Jaja, A. Rahman, Z. Omar, M. Zabidi, U. U. Sheikh","doi":"10.1109/ICSIPA.2017.8120582","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120582","url":null,"abstract":"This paper presents two intra prediction algorithms in High Efficiency Video Coding (HEVC) encoder for reducing the computational complexity and increase the encoding speed. The first algorithm takes advantage of the high spatial correlation among neighboring pixels to substitute the reference samples as the first row or first and second rows of the current block to be predicted, while the pixels intensities in the remaining rows or columns as in the case of horizontal predictions, are extrapolated as usual. Secondly, due to spatial correlations in video block data among adjacent blocks, it has been established that the intra prediction mode of the current block has a high probability of being a member of the most probable mode set or a slight variation of one of the most probable modes. These algorithms are combined and implemented on the HM16 reference software, and show speedup of 23.4% and 22.7% in encoding time using the all-intra-main configuration, with minimal reduction in bitrate of 0.21% and 0.22% respectively.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133310242","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-09-01DOI: 10.1109/ICSIPA.2017.8120642
Saleh Albelwi, A. Mahmood
Some of the best current face recognition approaches use feature extraction techniques based on either Principle Component Analysis (PCA), Local Binary Patterns (LBP), Autoencoder (non-linear PCA), etc. While each of these feature techniques works fairly well, we propose to combine multiple feature extractors with deep learning in a system so that the overall face recognition accuracy can be improved. The output from multiple feature extractions is classified using a deep learning approach. Deep learning algorithms possess high capability to learn more complex functions in order to handle difficult computer vison tasks. Our proposed method integrates the output of three different feature extractors, specifically PCA, LBP+PCA, and dimensionality reduction of LBP features using a Neural Network (NN). The features from the above three techniques are concatenated to form a joint feature vector. This feature vector is fed into a deep Sacked Sparse Autoencoder (SSA) as a classifier to generate the recognition results. Our proposed approach is evaluated by ORL and AR face databases. The experimental results indicate that our system outperforms existing ones based on individual feature techniques as well as reported systems employing multiple feature types.
{"title":"A deep architecture for face recognition based on multiple feature extraction techniques","authors":"Saleh Albelwi, A. Mahmood","doi":"10.1109/ICSIPA.2017.8120642","DOIUrl":"https://doi.org/10.1109/ICSIPA.2017.8120642","url":null,"abstract":"Some of the best current face recognition approaches use feature extraction techniques based on either Principle Component Analysis (PCA), Local Binary Patterns (LBP), Autoencoder (non-linear PCA), etc. While each of these feature techniques works fairly well, we propose to combine multiple feature extractors with deep learning in a system so that the overall face recognition accuracy can be improved. The output from multiple feature extractions is classified using a deep learning approach. Deep learning algorithms possess high capability to learn more complex functions in order to handle difficult computer vison tasks. Our proposed method integrates the output of three different feature extractors, specifically PCA, LBP+PCA, and dimensionality reduction of LBP features using a Neural Network (NN). The features from the above three techniques are concatenated to form a joint feature vector. This feature vector is fed into a deep Sacked Sparse Autoencoder (SSA) as a classifier to generate the recognition results. Our proposed approach is evaluated by ORL and AR face databases. The experimental results indicate that our system outperforms existing ones based on individual feature techniques as well as reported systems employing multiple feature types.","PeriodicalId":268112,"journal":{"name":"2017 IEEE International Conference on Signal and Image Processing Applications (ICSIPA)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114931147","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：