Search

E-commerce with the immersive scene of live streaming gives users a new interaction and consumption experience, extending a new path for traditional marketing. Live barrage, being the primary means of users’ social engagement in live e-commerce, encapsulates their inclination to purchase products, commonly referred to as consumption intention. The mining of live barrage enables the exploration of latent user behavior and consumption intention. The current methods employed for mining consumption intention in e-commerce live barrage encounter challenges including insufficient corpus, excessively brief text, and significant semantic deficiencies. Therefore, this study proposes a paradigm of consumption intention recognition by combining the text features of barrage: by constructing a dataset of Live barrage, cleaning the original dataset, and then using a fused BERT’s feature network model BERT-BiLSTM to identify the consumption intention in live e-commerce Live barrage. The experimental results demonstrate the efficacy of the proposed method for recognizing consumption intention. Finally, this study conducted text mining on live barrage containing consumption intent, including word cloud visualization, topic extraction, and co-occurrence analysis, further investigating consumers’ purchasing needs and intentions.

In this paper, a high-precision hot-pressed light guide plate defect detection model based on improved YOLOv7 is proposed. The model strengthens the spatial correlation between background and foreground by fusing global context information. A densely connected convolutional network is used to enhance the feature extraction capability and mitigate problems such as gradient vanishing while ensuring the maximum information flow in the network. Further, adaptive spatial feature fusion is used in the feature fusion structure of the model; the adaptive spatial feature fusion structure compensates for the small targets that are difficult to extract in high dimensions from low dimensions, thus solving the problem of detecting small targets that are easy to lose. Finally, a self-constructed dataset is built using images of hot-pressed light guide plates collected from industrial sites, and a large number of experiments are conducted. Experimental results show that the defect detection model has a mean average precision (mAP) of 99.1% and a detection speed of 127 FPS. Compared with the mainstream surface defect target detection algorithms, while ensuring the detection speed, the accuracy rate has been significantly improved, and the accuracy rate and real-time can meet the requirements of the industrial field inspection of hot-pressed light guide plate.

Graph Convolutional Network (GCN) has achieved significant success in many graph representation learning tasks. GCN usually learns graph representations by performing Neighbor Aggregation (NA) and Feature Transformation (FT) operations. Deep Adaptive Graph Neural Network (DAGNN) improves NA operation in the aggregation-transformation decoupled GCN, which enables the model to obtain a large receiver field. However, the problem with GCN is that when the model is trained to a deeper level, the performance decreases. In particular, the influence of NA and FT on model degradation has not been fully considered. In this work, we propose a new decoupled GCN architecture to enhance the performance of deep GCN. First, we conduct an experimental analysis of the impact of NA and FT operations on the degradation of the deep GCN model. Subsequently, we propose an Adaptive Aggregation-Transform Decoupled Graph Convolutional Network (AATD-GCN) which divides the model into two depths $D_{NA}$ and $D_{FT}$ , and proposes improved approaches in NA and FT, respectively. The AATD-GCN reduces the influence of NA and FT operations on performance degradation of deep GCN, and obtains a large receiver field while extracting complex feature information. We investigate the requirements of model depth $D_{NA}$ and $D_{FT}$ for graphs of different structures and sizes. Finally, the effectiveness of the proposed architecture is verified through extensive experiments on real-world datasets, the experimental results show that AATD-GCN is superior in terms of accuracy and robustness.

Path planning is a significant and challenging task in the domain of automatic driving. Many applications, such as autonomous driving, robotic navigation and aircraft object tracking in complex and changing urban road scenes, need accurate and robust path planning by detecting obstacles in the forward direction. The traditional methods only rely on the path search method without considering the environmental factors, the vehicle path planning method cannot deal with the complex and changeable environment. To deal with above problems, we propose a perception-aware based multi-modal feature fusion approach that combines visual-inertial odometer (VIO) poses and semantic obstacles in the forward scene of vehicles to plan driving paths. The proposed method takes environment awareness as the guide and combines path search algorithm to realize path optimization task in complex environment. The proposed approach first uses a long short memory network (LSTM) to build a VIO that fuses visual and inertial data for pose estimation. To detect obstacles, the proposed method uses a segmentation model with a lightweight structure to extract semantic 3D landmarks. Finally, a path search strategy combining an A* algorithm and visual information is proposed to plan driving paths for intelligent vehicles. We estimate the proposed path planning method on assimilated scenes and public datasets (KITTI and Cityscapes) by using a micro controller (Jetson Xavier NX) installed on a small vehicle. We also show comparable results with path planning that only uses the greedy algorithm or heuristic algorithm without using visual information and show that our method is adequate in coping with different complex scenes.

The United States of America (USA) has been the most powerful country in the world for decades. The country's global leadership and power are critical for international security and intercountry relationships, in which the political events yield differences. With the deepening development of globalization and the recovery of the global economy, the future policies and presidential parties of the USA should be predicted to avoid potential crises. However, making unbiased and robust predictions is challenging. This study presents a weight-and-structure-determination (WASD) algorithm-based feedforward neural network activated by a set of square wave functions to predict the presidential parties in the USA from 2021 to 2037. The historical data of presidential parties from 1853 to 2017 are used as the training set in the numerical experiments, and the results show that the proposed neural network can predict 9 terms of future presidential parties. Compared with other models, the proposed model is more efficient, robust, purely time-driven, and bias-independent. Through the square wave-activated WASD neural network, the Democratic Party is predicted to win in 2025, 2029, and 2033, whereas the Republican Party might win in 2021 and 2037.

MMC (Modular Multilevel Converter) is the preferred topology for renewable energy delivery, but renewable energy resources, such as wind farms, contains much harmonic components. Compared with two-level VSC (Voltage Source Converter), the internal dynamic characteristics of MMC have great impact on the harmonic transmission. This paper quantitatively analyzes the harmonic transferring mechanism from one side to the other side of MMC, and deduces the time-domain analytical expression of the harmonic response on the MMC's other side. Firstly, the influence of background harmonic on MMC's control system and electrical system are quantitatively analyzed. On this basis, the equivalent circuit of MMC's electrical system considering its internal dynamics is established to accurately calculate the effect of background harmonic on the MMC's other side. In addition, the harmonic transferring impedance in frequency-domain is also defined in this paper, based on that the influence of background harmonic on the MMC's other side can be calculated quickly. According to the harmonic transferring characteristics analyzed above, a zero-sequence virtual damping strategy is proposed for the first time to suppress the background harmonic transferring. Finally, a 232-level 500kV MMC simulated model is built in MATLAB/Simulink to verify the accuracy of the harmonic transferring calculation method and the effectiveness of the virtual damping suppression strategy.

The online time-variant matrix inversion problem has attracted extensive attention and study, because of its considerable appearance and application in scientific research and industrial production. For various control optimization problems, the demand for the high-precision and rapid-convergence of matrix inversion algorithm is increasing. It remains an ongoing challenge due to the rigorous requirements of precision and convergence of the algorithm. In this paper, on the basis of our previous works, by using Zhang neural network (ZNN) method, a continuous time-variant matrix inversion model, which is also a Getz-Marsden dynamic system (i.e., GMDS-ZNN model 3), is proposed. Besides, a general ten-instant Zhang et al discretization (ZeaD) formula is presented and investigated, with corresponding theoretical results being provided. Next, by applying this general formula to discretize the continuous time-variant matrix inversion model, a general ten-instant discrete time-variant matrix inversion (DTVMI) algorithm with the sixth-order precision is proposed. For comparison, four other DTVMI algorithms, with the second-, third-, fourth-, and fifth-order precisions, are also proposed and presented, respectively, by using other ZeaD formulas to discretize the continuous time-variant matrix inversion model. Besides, for the situation of coefficient matrix derivative being unknown, we provided the formula of estimating it with the fifth-order precision. With the help of the proposed matrix derivative estimation formula, the actual application field of GMDS-ZNN model 3 is expanded evidently. Finally, theoretical analyses and simulation experiment results highlight the effectiveness and accuracy of the proposed GMDS-ZNN model 3 and DTVMI algorithms.

For current steganalysis of the image content-adaptive steganography, there are multiple problems to be improved, such as high difficulty and low accuracy, when detecting images with various contents and textures. For this problem, an improved steganalytic method is proposed in this paper based on the pre-classification and feature selection. First, using the features extracted based on the dependency analysis of image adjacent data, the images with various content and texture complexities are pre-classified as multiple clusters by the K-means algorithms. Then, the performance of existing various steganalytic features are analyzed for different clusters of images, and the optimal features for each cluster are selected for final classification. The experimental results show that the detection accuracy could be improved by the proposed method, and the rationality and availability are also verified. At the same time, the analysis and experimental results in this paper also show that the images with rich content and complex texture should be paid more attention both in steganography and in steganalysis.

The PHase-Aware pRojection Model (PHARM) feature set, built as phase-aware histograms of quantized projections obtained by convolving residuals with random matrices, achieves competitive detection performance against modern adaptive JPEG steganography while having significant computational cost. In this paper, we propose three improvements to the original PHARM, making it more efficient and effective. First, we reduce the maximum projection matrix size to decrease the computational complexity of convolution and better capture steganographic embedding changes. Second, we select more than one phase pair per projection to compute phase-aware histograms, thus correspondingly reducing the number of projections for each residual. Third, the transposition symmetry is also taken into consideration to make our features more robust while preserving the feature dimensionality. The numerous experiments are given to demonstrate the efficiency and effectiveness of our improved PHARM.

In topography detection procedures for diamond wire, precision motion control is one of the key preconditions; the motion control unit of detection needs both micron resolutions and centimeter strokes. In this paper, a linear motor with four sets of multilayer piezoelectric actuators was developed to achieve an actuator with high resolution and a large stroke. The stroke depended on the guideways of the system, which mar enlarged the stroke of a single multilayer piezoelectric actuator. The mechanism of differential actuation was analyzed by dividing an operation cycle into several states. A symmetrically arranged four-foot design was proposed, and a corresponding excitation method was devised. A prototype was fabricated to conduct a proof-of-concept operation test. Experimental results validated the feasibility of the prototype with a 10 mm stroke. The speed of the prototype increased when the amplitudes of driving voltage signals increased from 30 V to 100 V. Within the frequency range of 1 Hz to 120 Hz, the speed increased initially then decreased with driving frequency, the maximum of which was at 100 Hz. With an excitation voltage signal of 100 V and 100 Hz, the speed reached a maximum of 740 μm/s. With an excitation voltage signal of 30 V and 120 Hz, the resolution of the prototype reached 0.26 μm. The prototype satisfies precision motion control of topography detection for diamond wire. This work also provides an option for automatic systems that require high resolutions and large strokes.

We would like to add funding information regarding our article [1] “This work was supported by the Fundamental Research Funds for the Central Universities under Grant 328201905.”

Currently, the surface topology at large domestic airports with increasing traffic flow is becoming more complex. Taxiing conflicts grow as well. This paper proposed a strategy based on aircraft priority to solve the problem. Under the premise of ensuring aircraft priority, this strategy resolved the problem reasonably and improved the safety and efficiency of airport operation. With the purpose of minimizing time cost of all aircraft, an aircraft taxiing model was built. Simulation results showed the advantages of this strategy. Compared with a common conflict resolution strategy, taxiing routes of 9 aircraft changed. The total waiting time was reduced by 50% and the running time was reduced by 43.6s. In addition, to ensure taxiing efficiency of high-priority aircraft, low-priority aircraft tended to choose more tortuous routes when choosing suboptimal alternative routes.

The pine wilt disease (PWD) is one of the most dangerous and destructive diseases to coniferous forests. The rapid spread trend and strong destruction directly threaten the security of forests. The complex spread pattern and the hard labor process of diagnosis call for an effective way to detect the infected areas. In this paper, an airborne edge-computing and lightweight deep learning based system are designed for PWD detection by using imagery sensors. Unmanned aerial vehicle (UAV) is firstly utilized to realize a large-scale coverage of forests, which can substantially reduce the hard labor. Except for infected trees, a large number of irrelevant images are also acquired by the UAV, which will overload the burden of process and transmission. Then a lightweight improved YOLOv4-Tiny based method (named as YOLOv4-Tiny-3Layers) is proposed to filter these uninterested images by leveraging the computation capability of edge computing, which can realize a fast coarse-grained detection with a low missing rate. Finally, all the remaining images are transmitted to the ground workstation for the final fine-grained detection. Experimental results show that the proposed system can implement a fast detection with superior performance as compared to other methods, which helps to detect the infected pine trees in a quick manner.

The characterization of micro cracks on metal surface plays an important role in the process of manufacturing and using, which has attracted a lot of attention. This is mainly due to the fact that the size of defects is smaller and the depth of defects is difficult to predict compared with metal materials with larger plane, which is still challenging. In order to solve this problem, this paper proposes a characterization method of surface defects based on reflective laser thermography, and designs a laser heating nondestructive testing system based on reflection. The system includes a semiconductor laser to heat the surface of metal cracks, and an infrared imager to record changes in the temperature field of the metal surface. In the process of data analysis, an Otsu adaptive threshold segmentation method is selected to quantify the defect size, which can control the quantification accuracy of defect size within 25%. A derivative analysis method is proposed to quantify the depth of defects, which can control the depth quantification accuracy of tiny defects within 7%.

As an important part of non-destructive testing, infrared thermography testing is widely used in various fields of industrial development for monitoring the quality of metal parts. Considering the problem of low detection rate of surface defects on steel parts, we explored the application of neural architecture search (NAS) in infrared thermography area for the first time. On the one hand, we compared different time-series temperature features of defect locations in infrared images and validate the performance of three different features such as heating, cooling and full process by machine learning methods. On the other hand, we searched for multilayer perceptron through NAS technology to classify defects with different depths. Experiments have proved that the time-series temperature feature is very effective when used in the depth classification of defects, and the accuracy rate can reach 93% under the verification of traditional machine learning methods. The NAS technique used in this paper can search 100 multilayer perceptrons in a minimum of 121s and achieve 100% defect classification accuracy.

Research on forecasting methods of time series data has become one of the hot spots. More and more time series data are produced in various fields. It provides data for the research of time series analysis method, and promotes the development of time series research. Due to the generation of highly complex and large-scale time series data, the construction of forecasting models for time series data brings greater challenges. The main challenges of time series modeling are high complexity of time series data, low accuracy and poor generalization ability of prediction model. This paper attempts to cover the existing modeling methods for time series data and classify them. In addition, we make comparisons between different methods and list some potential directions for time series forecasting.

In frequency-modulated continuous-wave (FMCW) radar interference suppression based on signal reconstruction, the pruned exact linear time (PELT) algorithm is used to detect the time positions of the interference. Due to the uncertain penalty factor of the PELT algorithm, the exactness of the position detection is reduced; thus, the suppression performance is degraded. We propose a PELT algorithm with a known change number (PELT-KCN), where a known change number is used to calculate the optimal penalty factor such that the high accuracy of the algorithm can be guaranteed. After interference recognition, the beat signal is separated into two parts: the undamaged signal and the damaged signal. The former is utilized to restore the latter through an autoregressive (AR) model. In simulations and field experiments, we applied our proposed PELT-KCN algorithm to the interference suppression method and verified its performance. Our method can accurately detect the time positions of interference and effectively improve the signal-to-noise ratio (SNR) of the detected targets.

The visible light communication (VLC) and wireless fidelity (WIFI) heterogeneous network is considered as a highly competitive candidate for the next generation indoor wireless communication. VLC provides high data rate as downlink transmission while WIFI can be supplied to achieve uplink transmission and make up for link interruptions caused by occlusion. Due to the dense distribution of VLC access points (APs), data rate of users is reduced by inter-cell interference (ICI) seriously, therefore, interference management (IM) is crucial to ensure user satisfaction. Meanwhile, as the number of users increases, in order to ensure the needs of users, reasonable resource allocation is required in such heterogeneous network. Hence, the center of the elite user clustering attracts its non-interfering users to join a cluster that is a user with good channel quality and high demand. And users in one cluster share resources by sub-channel reuse. In consideration of interference management and resource allocation, we investigated resource allocation based on elite users clustering for interference management in heterogeneous network, which is formulated as a throughput maximization problem, and a graph-based IM is proposed to solve it. Compared with the benchmark 1 and benchmark 2 on a rectangular layout and an irregular layout, the simulation results demonstrate that the throughput of the proposed elite user clustering algorithm is increased by 19.5% and 12%, the average user satisfaction ratio is increased by 14% and 9%, respectively.

Neural network-based approaches have become the driven forces for Natural Language Processing (NLP) tasks. Conventionally, there are two mainstream neural architectures for NLP tasks: the recurrent neural network (RNN) and the convolution neural network (ConvNet). RNNs are good at modeling long-term dependencies over input texts, but preclude parallel computation. ConvNets do not have memory capability and it has to model sequential data as un-ordered features. Therefore, ConvNets fail to learn sequential dependencies over the input texts, but it is able to carry out high-efficient parallel computation. As each neural architecture, such as RNN and ConvNets, has its own pro and con, integration of different architectures is assumed to be able to enrich the semantic representation of texts, thus enhance the performance of NLP tasks. However, few investigation explores the reconciliation of these seemingly incompatible architectures. To address this issue, we propose a hybrid architecture based on a novel hierarchical multi-granularity attention mechanism, named Multi-granularity Attention-based Hybrid Neural Network (MahNN). The attention mechanism is to assign different weights to different parts of the input sequence to increase the computation efficiency and performance of neural models. In MahNN, two types of attentions are introduced: the syntactical attention and the semantical attention. The syntactical attention computes the importance of the syntactic elements (such as words or sentence) at the lower symbolic level and the semantical attention is used to compute the importance of the embedded space dimension corresponding to the upper latent semantics. We adopt the text classification as an exemplifying way to illustrate the ability of MahNN to understand texts. The experimental results on a variety of datasets demonstrate that MahNN outperforms most of the state-of-the-arts for text classification.

Image denoising is a well-researched problem in the image processing field. Numerous image denoising algorithms have been proposed in the past. Although researchers have continually focused on improving the denoising algorithm performance regarding denoising effect and outstanding results have been achieved, the improvement amplitude of a single denoising algorithm has decreased over time. Recently, the consensus neural network (CsNet), which combines multiple image denoisers to produce an overall better result compared to single algorithms, was proposed. However, the denoising process of CsNet is time-consuming owing to its MSE optimal weight setting and iterative boosting stages. Therefore, we propose an improved combination of nonlocally centralized sparse representation (NCSR) with a fast and flexible denoising convolutional neural network (FFDNet) using a spatial local fusion strategy (ICID). ICID uses a structural-based patch to decompose their denoised images into the strength, structure, and mean intensity components. Thereafter, an image patch is reconstructed and placed back into the fused image after fusing the three components separately. Experimental results verified that our algorithm is superior to CsNet, and it is faster. The combination of NCSR and FFDNet can harmonize the complementary denoising capabilities of different denoising algorithms. NCSR can preserve as many details as possible in natural images with numerous repeated structures, whereas FFDNet can achieve state-of-the-art results with a sufficiently large training set of images. Moreover, ICID uses the structural-based method, which considers more local details and preserves more textures, resulting in superior performance.

A self-training algorithm is an iterative method for semi-supervised learning, which wraps around a base learner. It uses its own predictions to assign labels to unlabeled data. For a self-training algorithm, the classification ability of the base learner and the estimation of prediction confidence are very important. The classical decision tree as the base learner cannot be effective in a self-training algorithm, because it cannot correctly estimate its own predictions. In this paper, we propose a novel method of node split of the decision trees, which uses weighted features to cluster instances. This method is able to combine multiple numerical and categorical features to split nodes. The decision tree and random forest constructed by this method are called FWCDT and FWCRF respectively. FWCDT and FWCRF have the better classification ability than the classical decision trees and forests based on univariate split when the training instances are fewer, therefore, they are more suitable as the base classifiers in self-training. What's more, on the basis of the proposed node-split method, we also respectively explore the suitable prediction confidence measurements for FWCDT and FWCRF. Finally, the results of experiment implemented on the UCI datasets show that the self-training feature weighted clustering decision tree (ST-FWCDT) and random forest (ST-FWCRF) can effectively exploit unlabeled data, and the final obtained classifiers have better generalization ability.

Human activity, which usually consists of several actions (sub-activities), generally covers interactions among persons and/or objects. In particular, human actions involve certain spatial and temporal relationships, are the components of more complicated activity, and evolve dynamically over time. Therefore, the description of a single human action and the modeling of the evolution of successive human actions are two major issues in human activity recognition. In this paper, we develop a method for human activity recognition that tackles these two issues. In the proposed method, an activity is divided into several successive actions represented by spatio-temporal patterns, and the evolution of these actions are captured by a sequential model. A refined comprehensive spatio-temporal graph is utilized to represent a single action, which is a qualitative representation of a human action incorporating both the spatial and temporal relations of the participant objects. Next, a discrete hidden Markov model is applied to model the evolution of action sequences. Moreover, a fully automatic partition method is proposed to divide a long-term human activity video into several human actions based on variational objects and qualitative spatial relations. Finally, a hierarchical decomposition of the human body is introduced to obtain a discriminative representation for a single action. Experimental results on the Cornell Activity Dataset demonstrate the efficiency and effectiveness of the proposed approach, which will enable long videos of human activity to be better recognized.

Structural optimization of complex mechatronic components may involve heterogeneous competing performance indices, including the cost, fixation, benefit, and deviation ones. However, such optimization problems with heterogeneous objectives have not been investigated so far. In this paper, a novel interval heterogeneous-objective robust optimization approach is proposed for complex mechatronic components. First, a unified interval heterogeneous-objective robust optimization model is constructed for mechatronic components with the uncertainties described as interval variables. Subsequently, a new interval robust equilibrium optimization algorithm is proposed to solve the interval heterogeneous-objective robust optimization model. Specifically, the unified formulas for assessing the robustness of interval heterogeneous-objective performance indices and the robust equilibrium among them are derived at first. Then, the preferential guidelines considering the robust equilibrium among all the objective and constraint performance indices are proposed for the direct ranking of various design vectors, and finally, the heterogeneous-objective robust equilibrium optimization of complex mechatronic components is realized by integrating the Kriging technique and the nested genetic algorithm. The feasibility and effectiveness of the proposed heterogeneous-objective robust optimization approach are verified by a numerical example and a case study.

The problem of characterizing the sensitivity indices of structural performance considering random and interval uncertainty variables simultaneously is analyzed. Data-driven polynomial chaos expansion (PCE) is established to treat random variables with arbitrary input uncertainty information, and subsequently extended to consider interval uncertainty variables simultaneously. An adaptive-sparse polynomial chaos expansion calculation algorithm is proposed to determine the appropriate polynomial coefficients based on the available uncertainty information of random and interval uncertainty variables. Subsequently, the local and total sensitivity indices of random and interval uncertainty variables are calculated directly using the constructed PCE model. Two numerical and engineering examples are presented to demonstrate the effectiveness of the proposed approach. Compared with the results of the Monte Carlo simulation method, the proposed approach yields excellent results with significantly reduced computational effort in the three examples.

In this paper, an efficient hybrid reliability analysis (HRA) method and a hybrid reliability-based design optimization (HRBDO) approach are proposed for realistic complex engineering structures with random and interval uncertainties. First, the HRBDO model for complex engineering structures is constructed with its objective and performance functions described as the implicit functions of design variables and random and interval parameters. Then, an efficient HRA method based on adaptive step size (ASS-HRA) is put forward to calculate the minimum reliability of the structure's performance function under the influences of both random and interval uncertainties, the computational efficiency and accuracy of which are verified by a benchmark test. Subsequently, an efficient HRBDO approach integrating the proposed ASS-HRA method with the polynomial response surface model (PRSM) is developed for solving the HRBDO problems of complex engineering structures, the effectiveness of which is demonstrated by a numerical example. Finally, the HRBDO of a high-speed press slider demonstrates the efficiency, effectiveness, and versatility of the proposed HRBDO approach based on the ASS-HRA in the design of realistic complex engineering structures.

Neoadjuvant chemotherapy (NAC) has become the main treatment option for breast cancer. Its adverse drug reactions (ADRs) make NAC painful both physiologically and psychologically. The factor pathological complete remission (pCR) describes how well a series of six or more chemotherapeutic treatments works on a patient. This study investigated the possibility of predicting pCR using only the nodal sizes of the first three treatments. A best feature combination for each breast cancer subtype was screened from the real nodal sizes of the first three treatments and the nodal sizes` of the next three treatments predicted from those of the first three ones. The prediction was evaluated by the metrics Avc = (sensitivity + specificity)/2. A triple-negative breast cancer (TN) patient may have an estimation of pCR Avc = 0.8696 after taking just three treatments. At least Avc = 0.7594 was achieved for all the four breast cancer subtypes investigated in this study.

Typical still action recognition methods rely on human body part detection and object detection. However, current human body part detectors and object detectors are far from perfect, leading to a negative impact on subsequent spatial relation learning of human-object interactions (HOIs). Bag-of-features (BoF)based methods go beyond such modeling paradigms, but they do not achieve the state-of-the-art accuracies. In this paper, we propose two still action recognition methods that model HOI layouts by image hierarchical representation, rather than explicitly constructing HOI relations. The first method encodes a dense set of SIFT features using Fisher vectors, where an image is divided into increasingly fine regions with the spatial pyramid. The second method takes recent pretrained deep networks as feature execrators, where an image is divided into overlapped regions. The improvement effect of the hierarchical representation is proven by extensive comparison experiments. Our methods are very simple and easy-to-use, which remarkably outperform those BoF-based methods and complicated human-centric methods. To the best of our knowledge, our methods achieve the highest accuracies to date on the Sports, PPMI, and extended PPMI data sets.