Pub Date : 2017-10-01DOI: 10.1109/CISP-BMEI.2017.8302216
M. Šavc, V. Glaser, A. Holobar, I. Cikajlo, Z. Matjačić
This contribution compares performances of nonnegative matrix factorization and high-density surface electromyogram (EMG) decomposition on EMG signals recoded from forearm muscles of young healthy subjects. During the EMG measurements, subjects performed dynamic wrist extensions and flexions and universal haptic device robot was used to oppose their movements and to measure wrist kinematics and excreted muscle forces. Recoded EMG signals were independently decomposed by Convolution Kernel Compensation technique and by alternating least squares non-negative matrix factorization. The identified motor unit discharge patterns were summed into cumulative spike trains and compared with non-negative components in each measurement. The results demonstrated good match (average correlation coefficient of 0.92 ± 0.06), but several discrepancies between the identified components have also been observed. In particular, when limiting the time support of identified components to active EMG signal portions only, the average correlation coefficient dropped to 0.72 ±0.20.
{"title":"Comparison of non-negative matrix factorization and convolution kernel compensation in surface electromyograms of forearm muscles","authors":"M. Šavc, V. Glaser, A. Holobar, I. Cikajlo, Z. Matjačić","doi":"10.1109/CISP-BMEI.2017.8302216","DOIUrl":"https://doi.org/10.1109/CISP-BMEI.2017.8302216","url":null,"abstract":"This contribution compares performances of nonnegative matrix factorization and high-density surface electromyogram (EMG) decomposition on EMG signals recoded from forearm muscles of young healthy subjects. During the EMG measurements, subjects performed dynamic wrist extensions and flexions and universal haptic device robot was used to oppose their movements and to measure wrist kinematics and excreted muscle forces. Recoded EMG signals were independently decomposed by Convolution Kernel Compensation technique and by alternating least squares non-negative matrix factorization. The identified motor unit discharge patterns were summed into cumulative spike trains and compared with non-negative components in each measurement. The results demonstrated good match (average correlation coefficient of 0.92 ± 0.06), but several discrepancies between the identified components have also been observed. In particular, when limiting the time support of identified components to active EMG signal portions only, the average correlation coefficient dropped to 0.72 ±0.20.","PeriodicalId":6474,"journal":{"name":"2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"1 1","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2017-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76287933","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 : 2017-10-01DOI: 10.1109/CISP-BMEI.2017.8302327
Mariusz Mulka, Wojciech A. Lorkiewicz, R. Katarzyniak
This paper introduces a computational solution allowing an artificial system to organise large datasets into a set of known basic-level categories. Following cognitive computing paradigm we present an approach towards category-based internal organisation of cognitive agent's semantic memory. In particular, assuming a given set of basic-level categories (predefined or developed) we provide a concise introduction to two perceptron-based computational models allowing an artificial system to classify objects into basic-level categories. Utilising results from other disciplines (psychology, linguistics and cognitive science) we take advantage of the notion of cue validity and incorporate it as underlying weights of input features. Finally, using real bird species dataset we highlight simulation results of classification's precision and recall measures.
{"title":"Object classification using basic-level categories","authors":"Mariusz Mulka, Wojciech A. Lorkiewicz, R. Katarzyniak","doi":"10.1109/CISP-BMEI.2017.8302327","DOIUrl":"https://doi.org/10.1109/CISP-BMEI.2017.8302327","url":null,"abstract":"This paper introduces a computational solution allowing an artificial system to organise large datasets into a set of known basic-level categories. Following cognitive computing paradigm we present an approach towards category-based internal organisation of cognitive agent's semantic memory. In particular, assuming a given set of basic-level categories (predefined or developed) we provide a concise introduction to two perceptron-based computational models allowing an artificial system to classify objects into basic-level categories. Utilising results from other disciplines (psychology, linguistics and cognitive science) we take advantage of the notion of cue validity and incorporate it as underlying weights of input features. Finally, using real bird species dataset we highlight simulation results of classification's precision and recall measures.","PeriodicalId":6474,"journal":{"name":"2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"41 1","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2017-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73150666","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 : 2017-10-01DOI: 10.1109/CISP-BMEI.2017.8302180
Qiushi Meng, Zhaoyang Jin
Most of the online dynamic magnetic resonance imaging (dMRI) techniques are developed based on Cartesian trajectories. Recently, radial trajectories have been proposed to acquire image data for online dMRI. Compared with Cartesian trajectories, radial trajectories cover densely at k-space center and are more incoherent. When using compressed sensing technique to reconstruct dynamic images with under-sampling radial k-space data, the regridding procedure is employed, however it is usually time consuming and introduces numerical errors. In this study, a novel radial-like pseudo-polar (PP) trajectory was used for online dMRI. PP trajectory can avoid regridding and inverse-regridding operation by using a pseudopolar FFT (PPFFT) operation without interpolation. In the reconstructiongraphics processing unit (GPU) is used to further decrease the reconstruction time and achieve real-time online effect. In this simulation study, cardiac k-space dataset was fully acquired and using as a reference dataset. The PP trajectory was used to retrospectively under-sample k-space data with 12.5% and 25% coverage. The reconstruction results show that, the image quality of online dMRI based on PP under-sampling is higher than that of radial under-sampling based method. The reconstruction time was significantly shorten by using GPU acceleration, for the tested case, it is more than 20 times faster than the CPU computing.
{"title":"Online dynamic magnetic resonance imaging based on pseudo-polar sampling and GPU acceleration","authors":"Qiushi Meng, Zhaoyang Jin","doi":"10.1109/CISP-BMEI.2017.8302180","DOIUrl":"https://doi.org/10.1109/CISP-BMEI.2017.8302180","url":null,"abstract":"Most of the online dynamic magnetic resonance imaging (dMRI) techniques are developed based on Cartesian trajectories. Recently, radial trajectories have been proposed to acquire image data for online dMRI. Compared with Cartesian trajectories, radial trajectories cover densely at k-space center and are more incoherent. When using compressed sensing technique to reconstruct dynamic images with under-sampling radial k-space data, the regridding procedure is employed, however it is usually time consuming and introduces numerical errors. In this study, a novel radial-like pseudo-polar (PP) trajectory was used for online dMRI. PP trajectory can avoid regridding and inverse-regridding operation by using a pseudopolar FFT (PPFFT) operation without interpolation. In the reconstructiongraphics processing unit (GPU) is used to further decrease the reconstruction time and achieve real-time online effect. In this simulation study, cardiac k-space dataset was fully acquired and using as a reference dataset. The PP trajectory was used to retrospectively under-sample k-space data with 12.5% and 25% coverage. The reconstruction results show that, the image quality of online dMRI based on PP under-sampling is higher than that of radial under-sampling based method. The reconstruction time was significantly shorten by using GPU acceleration, for the tested case, it is more than 20 times faster than the CPU computing.","PeriodicalId":6474,"journal":{"name":"2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"137 1","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2017-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73229150","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 : 2017-10-01DOI: 10.1109/CISP-BMEI.2017.8302134
Mohammad Shohidul Islam, Dong-Kai Yang, Muhammad Abdul Alim Sikder
GNSS remote sensing can play an important rule for monitoring the earth's surface. For this why, this technique is proposed to retrieve some information from water surface. Specifically, scattered signals from the ocean surface can provide significant information of water surface and its current condition. In this paper, direct and scattered power waveforms of GEO Satellite of Beidou are mainly focused and analyzed. Due to delay of reflected signal, the waveforms pattern & characteristics are clearly different from direct signals. These power waveforms are determined for different coherent integration times, different wind speeds and different satellite elevation angles. Z-V scattering model and Tanos Elfouhaily wave spectrum model are utilized to present the power waveforms of reflected signals and these waveforms are compared with the power waveforms of reflected & direct signals of real data. It is observed that these waveforms are sensitive to integration time, wind speed and satellite elevation angle. It is also found that the waveforms are directly related to these three parameters (time, wind speed & elevation angle). As a result, the model waveforms and real data waveforms have performed a good matching each other.
{"title":"Power waveforms analysis of GNSS-R delay signals from beidou GEO satellite","authors":"Mohammad Shohidul Islam, Dong-Kai Yang, Muhammad Abdul Alim Sikder","doi":"10.1109/CISP-BMEI.2017.8302134","DOIUrl":"https://doi.org/10.1109/CISP-BMEI.2017.8302134","url":null,"abstract":"GNSS remote sensing can play an important rule for monitoring the earth's surface. For this why, this technique is proposed to retrieve some information from water surface. Specifically, scattered signals from the ocean surface can provide significant information of water surface and its current condition. In this paper, direct and scattered power waveforms of GEO Satellite of Beidou are mainly focused and analyzed. Due to delay of reflected signal, the waveforms pattern & characteristics are clearly different from direct signals. These power waveforms are determined for different coherent integration times, different wind speeds and different satellite elevation angles. Z-V scattering model and Tanos Elfouhaily wave spectrum model are utilized to present the power waveforms of reflected signals and these waveforms are compared with the power waveforms of reflected & direct signals of real data. It is observed that these waveforms are sensitive to integration time, wind speed and satellite elevation angle. It is also found that the waveforms are directly related to these three parameters (time, wind speed & elevation angle). As a result, the model waveforms and real data waveforms have performed a good matching each other.","PeriodicalId":6474,"journal":{"name":"2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"24 1","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2017-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74619347","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 : 2017-10-01DOI: 10.1109/CISP-BMEI.2017.8301961
Hong Zhang, Guodong Wang, Nan Wu, Guojia Hou, Zhimei Zhang
Image blind restoration has been a significant subject in various application fields. In the paper, we mainly studied the color image. In the process of converting color image into gray image will result in the loss of information because color image has different channels. In order to solve blind deconvolution of color image effectively, we present a method that estimates kernel result from three channels of color image directly based on multiscale framework. And then we ues the Multichannel Total Variation (MTV) model to protect image edges. By using normalized hyper Laplacian prior term, our method can converge to the real solution. The final clear image can be gotten. Although the MTV model will increase the complexity of computation, The correctness of algorithm and the feasibility of methods are proved by experiments.
{"title":"Single-color image motion deblurring using MTV model","authors":"Hong Zhang, Guodong Wang, Nan Wu, Guojia Hou, Zhimei Zhang","doi":"10.1109/CISP-BMEI.2017.8301961","DOIUrl":"https://doi.org/10.1109/CISP-BMEI.2017.8301961","url":null,"abstract":"Image blind restoration has been a significant subject in various application fields. In the paper, we mainly studied the color image. In the process of converting color image into gray image will result in the loss of information because color image has different channels. In order to solve blind deconvolution of color image effectively, we present a method that estimates kernel result from three channels of color image directly based on multiscale framework. And then we ues the Multichannel Total Variation (MTV) model to protect image edges. By using normalized hyper Laplacian prior term, our method can converge to the real solution. The final clear image can be gotten. Although the MTV model will increase the complexity of computation, The correctness of algorithm and the feasibility of methods are proved by experiments.","PeriodicalId":6474,"journal":{"name":"2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"8 1","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2017-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73980359","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}
Text representation is one of the fundamental problems in text analysis tasks. The key of text representation is to extract and express the semantic and syntax feature of texts. The order-sensitive sequence models based on neural networks have achieved great progress in text representation. Bidirectional Long Short-Term Memory (BiLSTM) Neural Networks, as an extension of Recurrent Neural Networks (RNN), not only can deal with variable-length texts, capture the long-term dependencies in texts, but also model the forward and backward sequence contexts. Moreover, typical neural networks, Convolutional Neural Networks (CNN), can extract more semantic and structural information from texts, because of their convolution and pooling operations. The paper proposes a hybrid model, which combines the BiLSTM with 2-dimensial convolution and 1-dimensial pooling operations. In other words, the model firstly captures the abstract representation vector of texts by the BiLSTM, and then extracts text semantic features by 2-dimensial convolutional and 1-dimensial pooling operations. Experiments on text classification tasks show that our method obtains preferable performances compared with the state-of-the-art models when applied on the MR1 sentence polarity dataset.
{"title":"A novel method of text representation on hybrid neural networks","authors":"Yanbu Guo, Chen Jin, Weihua Li, Chen Ji, Yuanye Fang, Yunhao Duan","doi":"10.1109/CISP-BMEI.2017.8302099","DOIUrl":"https://doi.org/10.1109/CISP-BMEI.2017.8302099","url":null,"abstract":"Text representation is one of the fundamental problems in text analysis tasks. The key of text representation is to extract and express the semantic and syntax feature of texts. The order-sensitive sequence models based on neural networks have achieved great progress in text representation. Bidirectional Long Short-Term Memory (BiLSTM) Neural Networks, as an extension of Recurrent Neural Networks (RNN), not only can deal with variable-length texts, capture the long-term dependencies in texts, but also model the forward and backward sequence contexts. Moreover, typical neural networks, Convolutional Neural Networks (CNN), can extract more semantic and structural information from texts, because of their convolution and pooling operations. The paper proposes a hybrid model, which combines the BiLSTM with 2-dimensial convolution and 1-dimensial pooling operations. In other words, the model firstly captures the abstract representation vector of texts by the BiLSTM, and then extracts text semantic features by 2-dimensial convolutional and 1-dimensial pooling operations. Experiments on text classification tasks show that our method obtains preferable performances compared with the state-of-the-art models when applied on the MR1 sentence polarity dataset.","PeriodicalId":6474,"journal":{"name":"2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"43 1","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2017-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74114597","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 : 2017-10-01DOI: 10.1109/CISP-BMEI.2017.8301951
Jing Liu, Ruijiao Liu, Jinlei Chen, Yajie Yang, Douli Ma
An improved image denoising algorithm based on block-matching and 3D collaborative filtering (BM3D) is proposed in this manuscript. Instead of using the same filtering model for all patches in an image, we employ two different nonlocal filtering models in edge and smooth regions, respectively. We realize it by using the nonlocal centralized sparse representation (NCSR) to capture both local sparsity of wavelet coefficients and nonlocal similarity of grouped blocks. Experimental results demonstrate that the proposed method outperforms several state-of-the-art denoising methods in terms of objective metrics and visual quality.
{"title":"Collaborative filtering denoising algorithm based on the nonlocal centralized sparse representation model","authors":"Jing Liu, Ruijiao Liu, Jinlei Chen, Yajie Yang, Douli Ma","doi":"10.1109/CISP-BMEI.2017.8301951","DOIUrl":"https://doi.org/10.1109/CISP-BMEI.2017.8301951","url":null,"abstract":"An improved image denoising algorithm based on block-matching and 3D collaborative filtering (BM3D) is proposed in this manuscript. Instead of using the same filtering model for all patches in an image, we employ two different nonlocal filtering models in edge and smooth regions, respectively. We realize it by using the nonlocal centralized sparse representation (NCSR) to capture both local sparsity of wavelet coefficients and nonlocal similarity of grouped blocks. Experimental results demonstrate that the proposed method outperforms several state-of-the-art denoising methods in terms of objective metrics and visual quality.","PeriodicalId":6474,"journal":{"name":"2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"84 1","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2017-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79358964","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 : 2017-10-01DOI: 10.1109/CISP-BMEI.2017.8301999
Yiliu Feng, Zhengfa Liang, Hengzhu Liu
Stereo matching plays an important role in many applications, such as Advanced Driver Assistance Systems, 3D reconstruction, navigation, etc. However it is still an open problem with many difficult. Most difficult are often occlusions, object boundaries, and low or repetitive textures. In this paper, we propose a method for processing the stereo matching problem. We propose an efficient convolutional neural network to measure how likely the two patches matched or not and use the similarity as their stereo matching cost. Then the cost is refined by stereo methods, such as semiglobal maching, subpixel interpolation, median filter, etc. Our architecture uses large image patches which makes the results more robust to texture-less or repetitive textures areas. We experiment our approach on the KITTI2015 dataset which obtain an error rate of 4.42% and only needs 0.8 second for each image pairs.
{"title":"Efficient deep learning for stereo matching with larger image patches","authors":"Yiliu Feng, Zhengfa Liang, Hengzhu Liu","doi":"10.1109/CISP-BMEI.2017.8301999","DOIUrl":"https://doi.org/10.1109/CISP-BMEI.2017.8301999","url":null,"abstract":"Stereo matching plays an important role in many applications, such as Advanced Driver Assistance Systems, 3D reconstruction, navigation, etc. However it is still an open problem with many difficult. Most difficult are often occlusions, object boundaries, and low or repetitive textures. In this paper, we propose a method for processing the stereo matching problem. We propose an efficient convolutional neural network to measure how likely the two patches matched or not and use the similarity as their stereo matching cost. Then the cost is refined by stereo methods, such as semiglobal maching, subpixel interpolation, median filter, etc. Our architecture uses large image patches which makes the results more robust to texture-less or repetitive textures areas. We experiment our approach on the KITTI2015 dataset which obtain an error rate of 4.42% and only needs 0.8 second for each image pairs.","PeriodicalId":6474,"journal":{"name":"2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"16 1","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2017-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84342981","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 : 2017-10-01DOI: 10.1109/CISP-BMEI.2017.8302098
Banghua Yang, Xu Han, Jianzhen Tang
Emotion recognition is a hot spot in advanced humancomputer interaction system, which is of great significance in artificial intelligence, health care, distance education, military field and so on. The paper builds a stacked autoencoder deep learning classification network consist of an input layer, two autoencoder hidden layers and a softmax classifier output layer based on SJTU Emotion EEG Dataset (SEED). Pretrain the first autoencoder employed L-BFGS to optimize the cost function. Then pretrain the second autoencoder with the output of first autoencoder. Finally send to the softmax classifier. Pretrain each autoencoder in forward propagation, then fine-tuning the whole network in back propagation. The well-trained network is used to classify three emotion states including happy, neural and grief. The raw inputs are differential entropy of EEG signal in five rhythmic frequencies band and the differential entropy of whole EEG signal. Fourteen experiments are performed with 5-fold cross validation, the average classification accuracy of three class emotion states is 59.6%, 66.27%, 71.97%, 78.48%, 82.56% and 85.5%. The result shows the higher frequency band differential entropy like Gamma band is more relative to emotion reaction.
{"title":"Three class emotions recognition based on deep learning using staked autoencoder","authors":"Banghua Yang, Xu Han, Jianzhen Tang","doi":"10.1109/CISP-BMEI.2017.8302098","DOIUrl":"https://doi.org/10.1109/CISP-BMEI.2017.8302098","url":null,"abstract":"Emotion recognition is a hot spot in advanced humancomputer interaction system, which is of great significance in artificial intelligence, health care, distance education, military field and so on. The paper builds a stacked autoencoder deep learning classification network consist of an input layer, two autoencoder hidden layers and a softmax classifier output layer based on SJTU Emotion EEG Dataset (SEED). Pretrain the first autoencoder employed L-BFGS to optimize the cost function. Then pretrain the second autoencoder with the output of first autoencoder. Finally send to the softmax classifier. Pretrain each autoencoder in forward propagation, then fine-tuning the whole network in back propagation. The well-trained network is used to classify three emotion states including happy, neural and grief. The raw inputs are differential entropy of EEG signal in five rhythmic frequencies band and the differential entropy of whole EEG signal. Fourteen experiments are performed with 5-fold cross validation, the average classification accuracy of three class emotion states is 59.6%, 66.27%, 71.97%, 78.48%, 82.56% and 85.5%. The result shows the higher frequency band differential entropy like Gamma band is more relative to emotion reaction.","PeriodicalId":6474,"journal":{"name":"2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"21 1","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2017-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84477602","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 : 2017-10-01DOI: 10.1109/CISP-BMEI.2017.8301986
Kaihua Tang, Xiao-Nan Hou, Zhiwen Shao, Lizhuang Ma
While traditional PIE (pose, illumination and expression) face variations have been well settled by latest methods, a new kind of variation, cross-age variation, is drawing attention from researchers. Most of the existing methods fail to maintain the effectiveness in real world applications that contain significant gap of age. Cross-age variation is caused by the shape deformation and texture changing of human faces while people getting old. It will result in tremendous intra-personal changes of face feature that deteriorate the performance of algorithms. This paper proposed a deep feature based framework for face retrieval problem. Our framework uses deep CNNs feature descriptor and two well designed post-processing methods to achieve age-invariance. To the best of our knowledge, this is the first deep feature based method in cross-age face retrieval problem. The deep CNNs model we use is firstly trained on traditional PIE datasets and then fine-tuned by cross-age dataset. The feature selection and projection post-processing we propose is also proved to be very effective in eliminating cross-age variation of deep CNNs feature. The experiments conducted on Cross-Age Celebrity Dataset (CACD), which is the largest public dataset containing cross-age variation, show that our framework outperforms previous state-of-the-art methods.
{"title":"Deep feature selection and projection for cross-age face retrieval","authors":"Kaihua Tang, Xiao-Nan Hou, Zhiwen Shao, Lizhuang Ma","doi":"10.1109/CISP-BMEI.2017.8301986","DOIUrl":"https://doi.org/10.1109/CISP-BMEI.2017.8301986","url":null,"abstract":"While traditional PIE (pose, illumination and expression) face variations have been well settled by latest methods, a new kind of variation, cross-age variation, is drawing attention from researchers. Most of the existing methods fail to maintain the effectiveness in real world applications that contain significant gap of age. Cross-age variation is caused by the shape deformation and texture changing of human faces while people getting old. It will result in tremendous intra-personal changes of face feature that deteriorate the performance of algorithms. This paper proposed a deep feature based framework for face retrieval problem. Our framework uses deep CNNs feature descriptor and two well designed post-processing methods to achieve age-invariance. To the best of our knowledge, this is the first deep feature based method in cross-age face retrieval problem. The deep CNNs model we use is firstly trained on traditional PIE datasets and then fine-tuned by cross-age dataset. The feature selection and projection post-processing we propose is also proved to be very effective in eliminating cross-age variation of deep CNNs feature. The experiments conducted on Cross-Age Celebrity Dataset (CACD), which is the largest public dataset containing cross-age variation, show that our framework outperforms previous state-of-the-art methods.","PeriodicalId":6474,"journal":{"name":"2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"52 1","pages":"1-7"},"PeriodicalIF":0.0,"publicationDate":"2017-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84581452","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：