Spatio-temporal Interest Points Research Articles

Projection transform on spatio-temporal context for action recognition

This paper discusses the task of human action recognition. This task is important to applications like video surveillance and video retrieval. Most of the existing local interest points based works on human action analysis, lost the information about spatio-temporal distribution of features and neglected the relationship between features and each defined actions. In this paper, through the analysis of feature distribution and their interactions over spatio-temporal domain, we propose a novel projection transform to take the two factors into account. A video sequence of human action in our perspective can be modeled by three types of features of spatio-temporal interest points: the global projection transform feature, the relative position distribution feature and the bag of visual words based feature. Then a new context-K-nearest-neighbor classifier is utilized to fuse them to form discriminative feature sets for action matching. In most of the case, our experiments have indicated that the novel method outperforms other previous published results on the Weizmann and KTH datasets.

Evaluating spatiotemporal interest point features for depth-based action recognition

Human action recognition has lots of real-world applications, such as natural user interface, virtual reality, intelligent surveillance, and gaming. However, it is still a very challenging problem. In action recognition using the visible light videos, the spatiotemporal interest point (STIP) based features are widely used with good performance. Recently, with the advance of depth imaging technology, a new modality has appeared for human action recognition. It is important to assess the performance and usefulness of the STIP features for action analysis on the new modality of 3D depth map. In this paper, we evaluate the spatiotemporal interest point (STIP) based features for depth-based action recognition. Different interest point detectors and descriptors are combined to form various STIP features. The bag-of-words representation and the SVM classifiers are used for action learning. Our comprehensive evaluation is conducted on four challenging 3D depth databases. Further, we use two schemes to refine the STIP features, one is to detect the interest points in RGB videos and apply to the aligned depth sequences, and the other is to use the human skeleton to remove irrelevant interest points. These refinements can help us have a deeper understanding of the STIP features on 3D depth data. Finally, we investigate a fusion of the best STIP features with the prevalent skeleton features, to present a complementary use of the STIP features for action recognition on 3D data. The fusion approach gives significantly higher accuracies than many state-of-the-art results.

Evaluation of Color Spatio-Temporal Interest Points for Human Action Recognition

This paper considers the recognition of realistic human actions in videos based on spatio-temporal interest points (STIPs). Existing STIP-based action recognition approaches operate on intensity representations of the image data. Because of this, these approaches are sensitive to disturbing photometric phenomena, such as shadows and highlights. In addition, valuable information is neglected by discarding chromaticity from the photometric representation. These issues are addressed by color STIPs. Color STIPs are multichannel reformulations of STIP detectors and descriptors, for which we consider a number of chromatic and invariant representations derived from the opponent color space. Color STIPs are shown to outperform their intensity-based counterparts on the challenging UCF sports, UCF11 and UCF50 action recognition benchmarks by more than 5% on average, where most of the gain is due to the multichannel descriptors. In addition, the results show that color STIPs are currently the single best low-level feature choice for STIP-based approaches to human action recognition.

Human Action Recognition Using Multi-Velocity STIPs and Motion Energy Orientation Histogram *

Local image features in space-time or spatio-temporal interest points provide compact and abstract representations of patterns in a video sequence. In this paper, we present a novel human action recognition method based on multi-velocity spatio-temporal interest points (MVSTIPs) and a novel local descriptor called motion energy (ME) orientation histogram (MEOH). The MVSTIP detection includes three steps: first, filtering video frames with multi-direction ME filters at different speeds to detect significant changes at the pixel level; thereafter, a surround suppression model is employed to rectify the ME deviation caused by the camera motion and complicated backgrounds (e.g., dynamic texture); finally, MVSTIPs are obtained with local maximum filters at multi-speeds. After detection, we develop MEOH descriptor to capture the motion features in local regions around interest points. The performance of the proposed method is evaluated on KTH, Weizmann, and UCF sports human action datasets. Results show that our method is robust to both simple and complex backgrounds and the method is superior to other methods that are based on local features.

An Action Recognition Model Based on the Bayesian Networks

In this paper, we propose a novel human action recognition method based on Bayesian network model and high-level semantic concept (human action attribute). Firstly, we extract 3D-SIFT descriptors for each spatio-temporal interest point in the videos. Secondly, the bag of words is used as the low-level feature to represent these videos. Finally, Bayesian networks related to action attributes are trained based on MAP (Maximum A Posterior Probability) to recognize human behavior. The experimental results show that the proposed model is effective on action recognition.

Action Recognition in Realistic Scenes via Local Spatio-temporal Representation

In this paper, we consider the problem of human action recognition in realistic scenes. Mainstream methods adopt local spatio-temporal patterns matching strategy that detect STIPs (Spatio-Temporal Interest Points) and compute features from raw video frames, and then classify the features into some predefined action categories. However, these researches have made certain assumptions seldom hold in the real-world environment such as small scale, without viewpoint changes and background motion. To obtain more effective detection and description for human action videos of complex scenes, we improve the performance of STIPs detection in realistic scenes by filtering out noises resulted from low resolution or cluttered background. In addition, we propose a new feature description method that uses 3D local self-similarities to compensate for motion estimation error due to camera motion. BoW (Bag-of-Words) model is applied to learn a codebook in training stage, and a nonlinear SVM classifier is used for action recognition finally. Experimental results show that our approach achieve superior performance on more challenge dataset (Youtube) in comparison to baseline method.

Robust human action recognition based on spatio-temporal descriptors and motion temporal templates

Human action recognition is a challenging area of research because of its various potential applications in visual surveillance. Integrating motion temporal templates with spatio-temporal interest points based appearance descriptor for action recognition is proposed in this paper. First, we model the background in a scene using Gaussian mixture model and extract the foreground objects and binary mask of region of interest in a scene. Then, for the binary mask of object, we constructed motion temporal templates using the motion history image (MHI) and motion energy image (MEI); for the foreground objects, we performed spatio-temporal interest points (STIPs) detector. Thirdly, human action representation is combined with three-dimensional Scale Invariant Feature Transform descriptor (3D SIFT) on STIPs and Hu moments extracted from MHI and MEI. Generalized Multiple Kernel Learning (GMKL) is adapted to classification the actions. To validate the proposed descriptor, we have conducted extensive experiments on the KTH and Weizmann action datasets. The results prove that the 3D SIFT offers complementary information to motion temporal templates, showing its effectiveness for action representation; moreover, the proposed method yields comparable results with the state-of-the-art methods.

Sparse representation for robust abnormality detection in crowded scenes

In crowded scenes, the extracted low-level features, such as optical flow or spatio-temporal interest point, are inevitably noisy and uncertainty. In this paper, we propose a fully unsupervised non-negative sparse coding based approach for abnormality event detection in crowded scenes, which is specifically tailored to cope with feature noisy and uncertainty. The abnormality of query sample is decided by the sparse reconstruction cost from an atomically learned event dictionary, which forms a sparse coding bases. In our algorithm, we formulate the task of dictionary learning as a non-negative matrix factorization (NMF) problem with a sparsity constraint. We take the robust Earth Mover's Distance (EMD), instead of traditional Euclidean distance, as distance metric reconstruction cost function. To reduce the computation complexity of EMD, an approximate EMD, namely wavelet EMD, is introduced and well combined into our approach, without losing performance. In addition, the combination of wavelet EMD with our approach guarantees the convexity of optimization in dictionary learning. To handle both local abnormality detection (LAD) and global abnormality detection, we adopt two different types of spatio-temporal basis. Experiments conducted on four public available datasets demonstrate the promising performance of our work against the state-of-the-art methods.

Multimodal assistive technologies for depression diagnosis and monitoring

Depression is a severe mental health disorder with high societal costs. Current clinical practice depends almost exclusively on self-report and clinical opinion, risking a range of subjective biases. The long-term goal of our research is to develop assistive technologies to support clinicians and sufferers in the diagnosis and monitoring of treatment progress in a timely and easily accessible format. In the first phase, we aim to develop a diagnostic aid using affective sensing approaches. This paper describes the progress to date and proposes a novel multimodal framework comprising of audio-video fusion for depression diagnosis. We exploit the proposition that the auditory and visual human communication complement each other, which is well-known in auditory-visual speech processing; we investigate this hypothesis for depression analysis. For the video data analysis, intra-facial muscle movements and the movements of the head and shoulders are analysed by computing spatio-temporal interest points. In addition, various audio features (fundamental frequency f0, loudness, intensity and mel-frequency cepstral coefficients) are computed. Next, a bag of visual features and a bag of audio features are generated separately. In this study, we compare fusion methods at feature level, score level and decision level. Experiments are performed on an age and gender matched clinical dataset of 30 patients and 30 healthy controls. The results from the multimodal experiments show the proposed framework’s effectiveness in depression analysis.

자세 예측을 이용한 효과적인 자세 기반 감정 동작 인식

인간의 동작 인식에 대한 이전 연구는 주로 관절체로 표현된 신체 움직임을 추적하고 분류하는데 초점을 맞춰 왔다. 이 방식들은 실제 이미지 사용 환경에서 신체 부위에 대한 정확한 분류가 필요하다는 점이 까다롭기 때문에 최근의 동작 인식 연구 동향은 시공간상의 관심 점과 같이 저수준의, 더 추상적인 외형특징을 이용하는 방식이 일반화되었다. 하지만 몇 년 사이 자세 예측 기술이 발전하면서 자세 기반 방식에 대한 시각을 재정립하는 것이 필요하다. 본 연구는 외형 기반 방식에서 저수준의 외형특징만으로 분류기를 학습시키는 것이 충분한지에 대한 문제를 제기하면서 자세 예측을 이용한 효과적인 자세기반 동작인식 방식을 제안하였다. 이를 위해 다양한 감정을 표현하는 동작 시나리오를 대상으로 외형 기반, 자세 기반 특징 및 두 가지 특징을 조합한 방식을 비교하였다. 실험 결과, 자세 예측을 이용한 자세 기반 방식이 저수준의 외형특징을 이용한 방식보다 감정 동작 분류 및 인식 성능이 더 나았으며 잡음 때문에 심하게 망가진 이미지의 감정 동작 인식에도 자세 예측을 이용한 자세기반의 방식이 효과적이었다. Early researches in human action recognition have focused on tracking and classifying articulated body motions. Such methods required accurate segmentation of body parts, which is a sticky task, particularly under realistic imaging conditions. Recent trends of work have become popular towards the use of more and low-level appearance features such as spatio-temporal interest points. Given the great progress in pose estimation over the past few years, redefined views about pose-based approach are needed. This paper addresses the issues of whether it is sufficient to train a classifier only on low-level appearance features in appearance approach and proposes effective pose-based approach with pose estimation for emotional action recognition. In order for these questions to be solved, we compare the performance of pose-based, appearance-based and its combination-based features respectively with respect to scenario of various emotional action recognition. The experiment results show that pose-based features outperform low-level appearance-based approach of features, even when heavily spoiled by noise, suggesting that pose-based approach with pose estimation is beneficial for the emotional action recognition.

Combining appearance and structural features for human action recognition

In this paper, we propose to integrate structural information with appearance features for human action recognition. In local representations based on detected spatio-temporal interest points (STIPs), the layout of STIPs carries important cues of motion structures in video sequences, and is assumed to contain complementary information to appearance features. We aim to incorporate structures into the description of STIPs by combing with appearance features for action representation.Based on the previous work of the 3D shape context, we present an optimised version of 3D shape context to encode the layout information of STIPs. By combining the proposed optimised 3D shape context with appearance descriptors, e.g., HOG3D and 3D gradients, we provide a more informative and discriminative description of STIPs for action classification. To validate the proposed descriptor, we have conducted extensive experiments on the KTH and the UCF YouTube datasets. The results prove that the optimised 3D shape context offers complementary information to appearance features, showing its effectiveness for action representation; moreover, the proposed descriptor yields comparable results with the state-of-the-art methods.

Selecting Effective and Discriminative Spatio-Temporal Interest Points for Recognizing Human Action

Many successful methods for recognizing human action are spatio-temporal interest point (STIP) based methods. Given a test video sequence, for a matching-based method using a voting mechanism, each test STIP casts a vote for each action class based on its mutual information with respect to the respective class, which is measured in terms of class likelihood probability. Therefore, two issues should be addressed to improve the accuracy of action recognition. First, effective STIPs in the training set must be selected as references for accurately estimating probability. Second, discriminative STIPs in the test set must be selected for voting. This work uses ε-nearest neighbors as effective STIPs for estimating the class probability and uses a variance filter for selecting discriminative STIPs. Experimental results verify that the proposed method is more accurate than existing action recognition methods.

Automatic Definition of Regions of Interest on Renal Scintigraphic Images

Renography is a radionuclide technique introduced in medicine since the early sixties in the daily practice. It allows measuring the function of each kidney separately in case of unilateral or bilateral uro-nephrological disease. A new system for analysis of Renography is presented in this paper. Our system aims to show a fully automated method for drawing the renal regions of interest based on a multi-agent system that incorporates spatio-temporal interest points detection on scintigraphic images by using the HOG3D descriptor to initialize agents. We used both types of agents, namely supervisor agents and explorator agents, they communicate among themselves and they inspire in their behavior from the Fast Marching method. Our system was tested on many real patients’ cases and we have obtained encouraging results. Clinical validation showed a constant agreement between the approach presented and manual segmentation by professionals in the nuclear medicine.

Large scale continuous visual event recognition using max-margin Hough transformation framework

In this paper we propose a novel method for continuous visual event recognition (CVER) on a large scale video dataset using max-margin Hough transformation framework. Due to high scalability, diverse real environmental state and wide scene variability direct application of action recognition/detection methods such as spatio-temporal interest point (STIP)-local feature based technique, on the whole dataset is practically infeasible. To address this problem, we apply a motion region extraction technique which is based on motion segmentation and region clustering to identify possible candidate “event of interest” as a preprocessing step. On these candidate regions a STIP detector is applied and local motion features are computed. For activity representation we use generalized Hough transform framework where each feature point casts a weighted vote for possible activity class centre. A max-margin frame work is applied to learn the feature codebook weight. For activity detection, peaks in the Hough voting space are taken into account and initial event hypothesis is generated using the spatio-temporal information of the participating STIPs. For event recognition a verification Support Vector Machine is used. An extensive evaluation on benchmark large scale video surveillance dataset (VIRAT) and as well on a small scale benchmark dataset (MSR) shows that the proposed method is applicable on a wide range of continuous visual event recognition applications having extremely challenging conditions.

A Local 3-D Motion Descriptor for Multi-View Human Action Recognition from 4-D Spatio-Temporal Interest Points

In this paper, we address the problem of human action recognition in reconstructed 3-D data acquired by multi-camera systems. We contribute to this field by introducing a novel 3-D action recognition approach based on detection of 4-D (3-D space <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">$+$</tex> </formula> time) spatio-temporal interest points (STIPs) and local description of 3-D motion features. STIPs are detected in multi-view images and extended to 4-D using 3-D reconstructions of the actors and pixel-to-vertex correspondences of the multi-camera setup. Local 3-D motion descriptors, histogram of optical 3-D flow (HOF3D), are extracted from estimated 3-D optical flow in the neighborhood of each 4-D STIP and made view-invariant. The local HOF3D descriptors are divided using 3-D spatial pyramids to capture and improve the discrimination between arm- and leg-based actions. Based on these pyramids of HOF3D descriptors we build a bag-of-words (BoW) vocabulary of human actions, which is compressed and classified using agglomerative information bottleneck (AIB) and support vector machines (SVMs), respectively. Experiments on the publicly available i3DPost and IXMAS datasets show promising state-of-the-art results and validate the performance and view-invariance of the approach.

Action Search by Example Using Randomized Visual Vocabularies

Because actions can be small video objects, it is a challenging problem to search for similar actions in crowded and dynamic scenes when a single query example is provided. We propose a fast action search method that can efficiently locate similar actions spatiotemporally. Both the query action and the video datasets are characterized by spatio-temporal interest points. Instead of using a unified visual vocabulary to index all interest points in the database, we propose randomized visual vocabularies to enable fast and robust interest point matching. To accelerate action localization, we have developed a coarse-to-fine video subvolume search scheme, which is several orders of magnitude faster than the existing spatio-temporal branch and bound search. Our experiments on cross-dataset action search show promising results when compared with the state of the arts. Additional experiments on a 5-h versatile video dataset validate the efficiency of our method, where an action search can be finished in just 37.6 s on a regular desktop machine.

Human action recognition based on boosted feature selection and naive Bayes nearest-neighbor classification

In this paper we propose a method of feature selection using the AdaBoost algorithm for action recognition. Instead of detecting spatio-temporal interest points and using a ‘bag of features’ approach, we use densely sampled descriptors, either 3D-SIFT or 3D-HOG, and select the most discriminative subset using the AdaBoost algorithm. We obtain maximal accuracy with just 200 of the 3217 possible raw 3D features from each video sequence. Using the extremely simple naive Bayes nearest-neighbor (NBNN) classifier with the most discriminative 3D-SIFT features, we obtain accuracies of: 92.7%, 99.4%, 92.3% and 38.1% on the KTH, Weizmann, IXMAS and HMDB51 datasets, respectively. We also observe that the errors are reasonably equitably distributed across the different action classes.

Learning semantic features for action recognition via diffusion maps

Efficient modeling of actions is critical for recognizing human actions. Recently, bag of video words (BoVW) representation, in which features computed around spatiotemporal interest points are quantized into video words based on their appearance similarity, has been widely and successfully explored. The performance of this representation however, is highly sensitive to two main factors: the granularity, and therefore, the size of vocabulary, and the space in which features and words are clustered, i.e., the distance measure between data points at different levels of the hierarchy. The goal of this paper is to propose a representation and learning framework that addresses both these limitations. We present a principled approach to learning a semantic vocabulary from a large amount of video words using Diffusion Maps embedding. As opposed to flat vocabularies used in traditional methods, we propose to exploit the hierarchical nature of feature vocabularies representative of human actions. Spatiotemporal features computed around interest points in videos form the lowest level of representation. Video words are then obtained by clustering those spatiotemporal features. Each video word is then represented by a vector of Pointwise Mutual Information (PMI) between that video word and training video clips, and is treated as a mid-level feature. At the highest level of the hierarchy, our goal is to further cluster the mid-level features, while exploiting semantically meaningful distance measures between them. We conjecture that the mid-level features produced by similar video sources (action classes) must lie on a certain manifold. To capture the relationship between these features, and retain it during clustering, we propose to use diffusion distance as a measure of similarity between them. The underlying idea is to embed the mid-level features into a lower-dimensional space, so as to construct a compact yet discriminative, high level vocabulary. Unlike some of the supervised vocabulary construction approaches and the unsupervised methods such as pLSA and LDA, Diffusion Maps can capture local relationship between the mid-level features on the manifold. We have tested our approach on diverse datasets and have obtained very promising results.

Selective spatio-temporal interest points

Recent progress in the field of human action recognition points towards the use of Spatio-Temporal Interest Points (STIPs) for local descriptor-based recognition strategies. In this paper, we present a novel approach for robust and selective STIP detection, by applying surround suppression combined with local and temporal constraints. This new method is significantly different from existing STIP detection techniques and improves the performance by detecting more repeatable, stable and distinctive STIPs for human actors, while suppressing unwanted background STIPs. For action representation we use a bag-of-video words (BoV) model of local N-jet features to build a vocabulary of visual-words. To this end, we introduce a novel vocabulary building strategy by combining spatial pyramid and vocabulary compression techniques, resulting in improved performance and efficiency. Action class specific Support Vector Machine (SVM) classifiers are trained for categorization of human actions. A comprehensive set of experiments on popular benchmark datasets (KTH and Weizmann), more challenging datasets of complex scenes with background clutter and camera motion (CVC and CMU), movie and YouTube video clips (Hollywood 2 and YouTube), and complex scenes with multiple actors (MSR I and Multi-KTH), validates our approach and show state-of-the-art performance. Due to the unavailability of ground truth action annotation data for the Multi-KTH dataset, we introduce an actor specific spatio-temporal clustering of STIPs to address the problem of automatic action annotation of multiple simultaneous actors. Additionally, we perform cross-data action recognition by training on source datasets (KTH and Weizmann) and testing on completely different and more challenging target datasets (CVC, CMU, MSR I and Multi-KTH). This documents the robustness of our proposed approach in the realistic scenario, using separate training and test datasets, which in general has been a shortcoming in the performance evaluation of human action recognition techniques.

Fast Action Detection via Discriminative Random Forest Voting and Top-K Subvolume Search

Multiclass action detection in complex scenes is a challenging problem because of cluttered backgrounds and the large intra-class variations in each type of actions. To achieve efficient and robust action detection, we characterize a video as a collection of spatio-temporal interest points, and locate actions via finding spatio-temporal video subvolumes of the highest mutual information score towards each action class. A random forest is constructed to efficiently generate discriminative votes from individual interest points, and a fast top-K subvolume search algorithm is developed to find all action instances in a single round of search. Without significantly degrading the performance, such a top-K search can be performed on down-sampled score volumes for more efficient localization. Experiments on a challenging MSR Action Dataset II validate the effectiveness of our proposed multiclass action detection method. The detection speed is several orders of magnitude faster than existing methods.