The creation of long melody sequences requires effective expression coherent musical structure. However, there is no clear representation Recent works on music generation have suggested various approaches to deal with the structural information music, but generating a full-song long-term structure remains challenge. In this paper, we propose MELONS, framework based graph which consists eight types bar-level relations. MELONS adopts multi-step method transformer-based networks by factoring...

The linear and angular accelerometers are widely used to measure the basic dynamic parameters in various aspects. However, current calibration methods for determining their sensitivities quite independent. This study investigates a monocular vision method with excitation generation device, which can simultaneously determine sensitivities. utilizes improved Lucy–Richardson edge enhancement line segment detection extraction improve accuracy. Comparison experiments conventional show that...

Online multi-object tracking (MOT) has broad applications in time-critical video analysis scenarios such as advanced driver-assistance systems (ADASs) and autonomous driving. In this paper, the proposed system aims at multiple vehicles front view of an onboard monocular camera. The vehicle detection probes are customized to generate high precision detection, which plays a basic role following tracking-by-detection method. A novel Siamese network with spatial pyramid pooling (SPP) layer is...

In this paper, the 3-D wavelet-fractal coding was used to compress hyperspectral remote sensing image. The classical eight kinds of affine transformations in 2-D fractal image compression were generalized nineteen for compression. Hyperspectral date cube first translated by wavelet and then applied lowest frequency subband. remaining coefficients higher subbands encoding SPIHT. We use eight-fork tree division algorithm improve matching accuracy reduce time coding. new method had been tested...

Crowd gathering detection plays an important role in security supervision of public areas. Existing image-processing-based methods are not robust for complex scenes, and deep-learning-based mainly focus on the design network, which ignores inner feature crowd action. To alleviate such problems, this work proposes a novel framework Detection Group Gathering (DGG) based counting method using deep learning approaches statistics to detect gathering. The DGG contains three parts, i.e., Detecting...

This paper considers a MIMO communication system that includes transmitter, legitimate receiver and an eavesdropper. In such scenario, the eavesdropper has knowledge of pilot utilizes contamination to reduce secrecy rate. Firstly, we analyze effects malicious on channel estimation. The more power is, eavesdropping performance enhances, by deducing rate its asymptotic expression in data transmission phase. Besides, optimal allocation scheme for full-duplex simultaneously transmit noise...

Point cloud semantic segmentation predicts the class of each point, which can help AI machines perceive real 3D world. Recently, precision for large-scale point is limited by complex scenarios, data occlusion, and massive data, remains. This paper proposes neural radiance field convolution to analysis. First, conquer some existing methods represent local space as relative position without considering rotation property cloud. A new spatial direction representation called seven-dimensional...

Carbonate rock is one of the major exploration targets in Tarim Basin. The carbonate reservoirs are usually irregular fractured-vuggy bodies, and respond as "beaded" reflections seismic section. bead-like features matched with high-yield stable production wells excellently. Reservoirs similar strong confirmed different degrees lithology filling, some even fully filled, causing drilling to fail. It very important identify filled beads fine reservoir development. Previous abnormal geobody...

Recently, solving the crowd counting problem under occlusion and complex perspective is a hot but difficult topic. Existing methods mainly constructed counters in parallel perspective, when facing such as influences of height difference heavy occlusions, they fail to get good accuracy. To alleviate these problems, this work proposes novel interesting framework NOOMP (Need Only One More Point) for adaptation task nature scenes. Firstly, considers that common scenes our daily life usually have...

The detection of pedestrian which has been widely used in digital surveillance systems is a popular topic computer vision. This paper mainly discusses system video sequences captured from stationary camera hanging public scene. We describe an efficient combining background subtraction based on Gaussian Mixture Model (GMM) and object classification Histograms Oriented Gradients (HOG). first process moving objects segmentation using GMM. Then HOG detector to classify the into person...

Timbre morphing is a signal processing technology that involves the use of an interpolation algorithm to gradually change timbre one instrument into another. However, prepared target audio, which possesses same music content (such as score and rhythm) original essential input used in technique. To meet application requirements non-parallel many-to-one processing, we combined spectral feature with deep learning autoencoder provide generalization ability. The complex framework, focuses on...

In the computer vision field, understanding human dynamics is not only a great challenge but also very meaningful work, which plays an indispensable role in public safety. Despite complexity of dynamics, physicists have found that pedestrian motion crowd governed by some internal rules, can be formulated as model, and effective model importance for reconstructing various scenes. this paper, we revisit related research social psychology propose two-part based on shortest path principle. One...

In order to achieve a high compression ratio on hyperspectrum image, in this paper, an algorithm about 3-D fractal coding is proposed. This based theory and eight fork tree division. We expand the kinds of transform nineteen space, which improves matching accuracy, accelerate encoding process by using eight-fork division cube block classification. A simulation was done MATLAB 7.1, experimental results demonstrate has some practical sense.

This paper presents a new and efficient calibration method for line structured-light plane expand it to the multi-line structured light case.This only uses checkerboard as planar target, can move freely long structure is projected onto checkerboard.Combining process with Zhang's of camera calibration,we get plenty points which are in fit plane.To be more efficient, this propose project multiple planes projector .After two ,A frame established at intersection planes, then all pose deduced....

A novel fusion HOG descriptor and Faster-RCNN (HFaster-RCNN) for pedestrian detection method is proposed. The proposed generates fine candidate regions using saliency map region proposal network (Saliency RPN). In order to synthesize the complementary advantages of traditional hand-crafted feature deep learning, we realized classifier decision information by Bayesian posterior probabilities. Experimental results show that has better hit rate compared with single extraction method....

The Economic Lot Scheduling Problem (ELSP) has been well-researched for over half a century. This paper, by using extended basic period approach, offers new algorithm which differs from previous algorithms in that every item is of different importance. Such an item, whose daily cost very high, considered to be greater importance and called “key item”; such total production time per times long, “obstacle item”. By making key items’ multipliers good compatibility loading obstacle items...

Long-term melody generation may encounter the challenges such as inadequate melodic variation, resulting in monotony or unreasonable variation. In this work, we introduce time series prediction and propose a method of Music-FED to generate more creative harmonic melodies. The proposed approach adopts first-order difference describe relative motion, designs temporal music representation that makes model easily aware hierarchy notes. It then learns distribution motion variation with...