Search

This paper proposes a novel 3D action recognition technique that uses time-series information extracted from depth image sequences for use in systems of human daily activity monitoring. To this end, each action is represented as a multi-dimensional time series, where each dimension represents the position variation of one skeleton joint over time. The time series is then mapped onto a vector space using Dynamic Time Warping (DTW) distance. Furthermore, to employ the correlation-distinctiveness relationship of the sequences in recognition, this vector space is remapped onto a discriminative space using the regularized Fisher method, where final decisions about the actions are made. Unlike other available methods, the time-warping used in the mapping strategy makes the feature space robust to temporal variations of the motion sequences. Moreover, our method eliminates the need for a complicated design method for extracting the static and dynamic information of a motion sequence. Furthermore, most existing methods treat all skeletal joints identically for different actions, while some joints are more discriminative to distinguish a specific action. Thanks to the nature of the proposed features, we propose to use a separate set of discriminative joints, called joint importance map for each class of action. Evaluation results on four well-known datasets, TST, UTKinect, UCFKinect, and NTU RGB+D show competitive performance with the state-of-the-art methods in human action recognition.

In this paper, we present a machine learning model to estimate the blood pressure (BP) of a person using only his photoplethysmogram (PPG) signal. We propose algorithms to better detect some critical points of the PPG signal, such as systolic and diastolic peaks, dicrotic notch and inflection point. These algorithms are applicable to different PPG signal morphologies and improve the precision of feature extraction. We show that the logarithm of dicrotic notch reflection index, the ratio of low- to high-frequency components of heart rate (HR) variability signal, and the product of HR multiplied by the modified Normalized Pulse Volume (mNPV) are the key features in accurately estimating the BP using PPG signal. Our proposed method has achieved higher accuracies in estimating BP compared to the previously reported methods that only use PPG signal. For the systolic BP, the achieved correlation coefficient between the estimated values and the real values is 0.78, the mean absolute error of the estimated values is 8.22 mmHg, and their standard deviation is 10.38 mmHg. For the diastolic BP, the achieved correlation coefficient between the estimated and the real values is 0.72, the mean absolute error of the estimated values is 4.17 mmHg, and their standard deviation is 4.22 mmHg. The achieved results fall within Grade A for diastolic, Grade C for systolic and Grade B for mean BP based on BHS standard.

Introduction: B lymphocyte-induced maturation protein 1 (BLIMP1) encoded by the positive regulatory domain 1 gene (PRDM1), is a key regulator in T cell differentiation in mouse models. BLIMP1-deficiency results in a lower effector phenotype and a higher memory phenotype. Methods: In this study, we aimed to determine the role of transcription factor BLIMP1 in human T cell differentiation. Specifically, we investigated the role of BLIMP1 in memory differentiation and exhaustion of human T cells. We used CRISPR interference (CRISPRi) to knock-down BLIMP1 and investigated the differential expressions of T cell memory and exhaustion markers in BLIMP1-deficient T cells in comparison with BLIMP1-sufficient ex vivo expanded human T cells. Results: BLIMP1-deficiency caused an increase in central memory (CM) T cells and a decrease in effector memory (EM) T cells. There was a decrease in the amount of TIM3 exhaustion marker expression in BLIMP1-deficient T cells; however, there was an increase in PD1 exhaustion marker expression in BLIMP1-deficient T cells compared with BLIMP1-sufficient T cells. Conclusion: Our study provides the first functional evidence of the impact of BLIMP1 on the regulation of human T cell memory and exhaustion phenotype. These findings suggest that BLIMP1 may be a promising target to improve the immune response in adoptive T cell therapy settings.

Temporal Action Localization (TAL) task which is to predict the start and end of each action in a video along with the class label of the action has numerous applications in the real world. But due to the complexity of this task, acceptable accuracy rates have not been achieved yet, whereas this is not the case regarding the action recognition task. In this paper, we propose a new network based on Gated Recurrent Unit (GRU) and two novel post-processing methods for TAL task. Specifically, we propose a new design for the output layer of the conventionally GRU resulting in the so-called GRU-Split network. Moreover, linear interpolation is used to generate the action proposals with precise start and end times. Finally, to rank the generated proposals appropriately, we use a Learn to Rank (LTR) approach. We evaluated the performance of the proposed method on Thumos14 and ActivityNet-1.3 datasets. Results show the superiority of the performance of the proposed method compared to state-of-the-art. Specifically in the mean Average Precision (mAP) metric at Intersection over Union (IoU) of 0.7 on Thumos14, we get 27.52% accuracy which is 5.12% better than that of state-of-the-art methods., 10 pages, 6 figures

In Iranian Sign Language (ISL), alongside the movement of fingers/arms, the dynamic movement of lips is also essential to perform/recognize a sign completely and correctly. In a follow up of our previous studies in empowering the RASA social robot to interact with individuals with hearing problems via sign language, we have proposed two automated lip-reading systems based on DNN architectures, a CNN-LSTM and a 3D-CNN, on the robotic system to recognize OuluVS2 database words. In the first network, CNN was used to extract static features, and LSTM was used to model temporal dynamics. In the second one, a 3D-CNN network was used to extract appropriate visual and temporal features from the videos. The accuracy rate of 89.44% and 86.39% were obtained for the presented CNN-LSTM and 3D-CNN networks, respectively; which were fairly promising for our automated lip-reading robotic system. Although the proposed non-complex networks did not provide the highest accuracy for this database (based on the literature), 1) they were able to provide better results than some of the more complex and even pre-trained networks in the literature, 2) they are trained very fast, and 3) they are quite appropriate and acceptable for the robotic system during Human-Robot Interactions (HRI) via sign language.

We introduce a novel Regularized Kernel Projection Pursuit Regression method which is a two-step nonlinearity encoding algorithm tailored for such very low dimensional problems as fatigue detection. This way, the data nonlinearity can be investigated from two different perspectives, first by transforming the data into a high dimensional intermediate space and then by using their spline estimations to the output variables which allows for a hierarchical unfolding of data. Experimental results on the SEED database shows an average RMSE value of 0.1080% and 0.1054% respectively for the temporal and posterior areas of the brain. Our method is also validated by conducting some experiments on Parkinson's disease prediction which further demonstrate the efficiency of our method for low-dimensional regression problems.Traditional off-the-shelf regression methods like SVR, KSVR, and GLM methods all require their link functions to be previously selected which limits their effectiveness for encoding the nonlinearity of a highly complex low dimensional data set. Moreover, conventional PPR does not deal with the very low dimensionality of data. This paper proposes a novel regression algorithm to address the encoding problem of a highly complex low dimensional data, which is usually encountered in bio-neurological prediction tasks like EEG based driving fatigue detection.

This paper proposes a new method for view-invariant action recognition that utilizes the temporal position of skeletal joints obtained by Kinect sensor. In this method, the actions are represented as sequences of several pre-defined poses. After pre-processing, which includes skeleton alignment and scaling, the appropriate feature vectors are obtained for recognizing and discriminating the pose of every frame by the proposed Fisherposes method. The proposed regularized Mahalanobis distance metric is used in order to recognize both the involuntary and highly made-up actions at the same time. Hidden Markov model (HMM) is then used to classify the action related to an input sequence of poses. For taking into account the motion in the actions which are not separable by solely their temporal poses, histograms of trajectories are also proposed. The proposed action recognition method is capable of recognizing both the voluntary and involuntary actions, as well as pose-based and trajectory-based ones with a high accuracy rate. The effectiveness of the proposed method is experimented on three publicly available data sets, TST fall detection, UTKinect, and UCFKinect data sets.

Goal : Continuous blood pressure (BP) monitoring can provide invaluable information about individuals’ health conditions. However, BP is conventionally measured using inconvenient cuff-based instruments, which prevents continuous BP monitoring. This paper presents an efficient algorithm, based on the pulse arrival time (PAT), for the continuous and cuffless estimation of the systolic BP, diastolic blood pressure (DBP), and mean arterial pressure (MAP) values. Methods : The proposed framework estimates the BP values through processing vital signals and extracting two types of features, which are based on either physiological parameters or whole-based representation of vital signals. Finally, the regression algorithms are employed for the BP estimation. Although the proposed algorithm works reliably without any need for calibration, an optional calibration procedure is also suggested, which can improve the system's accuracy even further. Results : The proposed method is evaluated on about a thousand subjects using the Association for the Advancement of Medical Instrumentation (AAMI) and the British Hypertension Society (BHS) standards. The method complies with the AAMI standard in the estimation of DBP and MAP values. Regarding the BHS protocol, the results achieve grade A for the estimation of DBP and grade B for the estimation of MAP. Conclusion : We conclude that by using the PAT in combination with informative features from the vital signals, the BP can be accurately and reliably estimated in a noninvasive fashion. Significance : The results indicate that the proposed algorithm for the cuffless estimation of the BP can potentially enable mobile health-care gadgets to monitor the BP continuously.

Despite many advances in optical character recognition in general, there are still serious challenges remaining in recognizing Farsi text. The main reason is the cursive nature of the letters in written Farsi, i.e., depending on the position of a letter within a word, it might join to its neighboring letters, which consequently changes the shape of the character. As a result, each letter can have up to four different character shapes. In addition to the problem of segmenting the characters, the increased number of characters makes the recognition task even more challenging. This paper introduces a complete framework for character recognition, including a method for segmenting the characters and one for classifying the resulting separated characters. Character segmentation is performed using a new sliding-window algorithm with a high accuracy rate of 98.23%. With a total of 32 Farsi letters resulting in 114 character shapes, an almost perfect character recognition rate of 99.94% is achieved using the proposed Fisher characters method. The final system, including segmentation and recognition modules, achieves a recognition rate of 98.17% and is robust against the scale and rotation of the image, and the font size of the written text.

Developing time-invariant solutions for recognition of human action is still an important and open challenge. Three issues make time-invariant solutions so important: different speed of performing the same action by different people, latency in doing the actions and the existence of redundant frames in the recorded video. To overcome these problems, we propose a method based on the so-called memory of the joints to remember only the cumulative positive and negative movement of each joint. Hence, we transform action recognition from time-space to shape-space and the action recognition becomes the problem of shape classification. These shape features contain highly discriminative information and are robust against execution rate. After extracting these features, convolutional neural networks are used to classify the actions. UTKinect and UTD_MHAD skeleton dataset are used for testing this method and the results demonstrate the effectiveness of the proposed method.

In holistic-based 3D face recognition methods, which have been shown to be more promising than feature-based methods, the most commonly used feature for recognition is the 3D coordinate of the face points. According to the experiments in this work, surface normal vectors alone have more discriminative information than the coordinates, and utilizing them along with the coordinates of the points improves the recognition. However, because of the variation in the aspect ratio of the face of different individuals, registering the points of a face all together to the reference face does not result in an appropriate correspondence between their points. This outcome in particular affects the quality of the extracted surface normal vectors and consequently degrades the recognition performance. In this paper, it has been shown that by partitioning the point cloud of a face into smaller parts and then registering each part separately to the reference face, the recognition performance can be significantly improved.

Alternative medicine can be used to achieve the healing effects of the experimental medicine, it also can predict diseases and prevent them, but validity of them is unproven. Iridology is an alternative medicine that claims to predict tissue weaknesses in the body by looking at the iris. The main object of this paper is to validate the using of iridology to predict diabetes, to do so iris images of 106 diabetic patients and 124 healthy controls were obtained and evaluated. The designed method is completely automatic and independent of iridologists. Also a novel algorithm is developed to improve the efficiency by finding the best region of iris automatically. The results showed the accuracy of 91.8% in the best setting, which shows the effectiveness of the proposed method.

Recent years have seen an increasing trend in developing 3D action recognition methods. However, despite the advances, existing models still suffer from some major drawbacks including the lack of any provision for recognizing action sequences with some missing frames. This significantly hampers the applicability of these methods for online scenarios, where only an initial part of sequences are already provided. In this paper, we introduce a novel sequence-to-sequence representation-based algorithm in which a query sample is characterized using a collaborative frame representation of all the training sequences. This way, an optimal classifier is tailored for the existing frames of each query sample, making the model robust to the effect of missing frames in sequences (e.g. in online scenarios). Moreover, due to the collaborative nature of the representation, it implicitly handles the problem of varying styles during the course of activities. Experimental results on three publicly available databases, UTKinect, TST fall, and UTD-MHAD, respectively, show 95.48%, 90.91%, and 91.67% accuracy when using the beginning 75% portion of query sequences and 84.42%, 60.98%, and 87.27% accuracy for their initial 50%.

Human action recognition from skeletal data is one of the most popular topics in computer vision which has been widely studied in the literature, occasionally with some very promising results. However, being supervised, most of the existing methods suffer from two major drawbacks; (1) too much reliance on massive labeled data and (2) high sensitivity to outliers, which in turn hinder their applications in such real-world scenarios as recognizing long-term and complex movements. In this paper, we propose a novel unsupervised 3D action recognition method called Sparseness Embedding in which the spatiotemporal representation of action sequences is nonlinearly projected into an unwarped feature representation medium, where unlike the original curved space, one can easily apply the Euclidean metrics. Our strategy can simultaneously integrate the characteristics of nonlinearity, sparsity, and space curvature of sequences into a single objective function, leading to a more robust and highly compact representation of discriminative attributes without any need to label information. Moreover, we propose a joint learning strategy for dealing with the heterogeneity of the temporal and spatial characteristics of action sequences. A set of extensive experiments on six publicly available databases, including UTKinect, TST fall, UTD-MHAD, CMU, Berkeley MHAD, and NTU RGB+D demonstrates the superiority of our method compared with the state-of-the-art algorithms.

Brain Computer Interface (BCI) systems, which are based on motor imagery, enable human to command artificial peripherals by merely thinking to the task. There is a tremendous interest in implementing BCIs on portable platforms, such as Field Programmable Gate Arrays (FPGAs) due to their low-cost, low-power and portability characteristics. This article presents the design and implementation of a Brain Computer Interface (BCI) system based on motor imagery on a Virtex-6 FPGA. In order to design an accurate algorithm, the proposed method avails statistical learning methods such as Mutual Information (MI), Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM). It also uses Separable Common Spatio Spectral Pattern (SCSSP) method in order to extract features. Simulation results prove achieved performances of 73.54% for BCI Competition III-dataset V, 67.2% for BCI Competition IV-dataset 2a with all four classes, 80.55% for BCI Competition IV-dataset 2a with the first two classes, and 81.9% for captured signals. Moreover, the final reported hardware resources determine its efficiency as a result of using retiming and folding techniques from the VLSI architecture perspective. The complete proposed BCI system not only achieves excellent recognition accuracy but also remarkable implementation efficiency in terms of portability, power, time, and cost.