Real-time Processing Rates Research Articles

Wearable Online Freezing of Gait Detection and Cueing System.

This paper presents a real-time wearable system designed to assist Parkinson's disease patients experiencing freezing of gait episodes. The system utilizes advanced machine learning models, including convolutional and recurrent neural networks, enhanced with past sample data preprocessing to achieve high accuracy, efficiency, and robustness. By continuously monitoring gait patterns, the system provides timely interventions, improving mobility and reducing the impact of freezing episodes. This paper explores the implementation of a CNN+RNN+PS machine learning model on a microcontroller-based device. The device operates at a real-time processing rate of 40 Hz and is deployed in practical settings to provide 'on demand' vibratory stimulation to patients. This paper examines the system's ability to operate with minimal latency, achieving an average detection delay of just 261 milliseconds and a freezing of gait detection accuracy of 95.1%. While patients received on-demand stimulation, the system's effectiveness was assessed by decreasing the average duration of freezing of gait episodes by 45%. These preliminarily results underscore the potential of personalized, real-time feedback systems in enhancing the quality of life and rehabilitation outcomes for patients with movement disorders.

Dense 3D Point Cloud Environmental Mapping Using Millimeter-Wave Radar

To address the challenges of sparse point clouds in current MIMO millimeter-wave radar environmental mapping, this paper proposes a dense 3D millimeter-wave radar point cloud environmental mapping algorithm. In the preprocessing phase, a radar SLAM-based approach is introduced to construct local submaps, which replaces the direct use of radar point cloud frames. This not only reduces data dimensionality but also enables the proposed method to handle scenarios involving vehicle motion with varying speeds. Building on this, a 3D-RadarHR cross-modal learning network is proposed, which uses LiDAR as the target output to train the radar submaps, thereby generating a dense millimeter-wave radar point cloud map. Experimental results across multiple scenarios, including outdoor environments and underground tunnels, demonstrate that the proposed method can increase the point cloud density of millimeter-wave radar environmental maps by over 50 times, with a point cloud accuracy better than 0.1 m. Compared to existing algorithms, the proposed method achieves superior environmental map reconstruction performance while maintaining a real-time processing rate of 15 Hz.

Real-Time Query Method Based on Distributed Database Demand Information

This paper proposes a real-time query method for distributed database requirement information. First, we propose to extract the database information according to the submitted query keyword. Then, we establish a real-time model of distributed database demand information query and a real-time information classification model, using a distributed multivariate hash algorithm model and a query of real-time information. When there is a query keyword, the proposed method directly reads the data above the hash ring based on the keyword to improve query efficiency. Extensive experimental results indicate that our proposed method reduces the number of queries, query pressure, and processing time of information queries in the distributed database. Moreover, the distributed database’s real-time processing rate and amount have increased.

Dynamic Load Balancing Algorithm Based on Optimal Matching of Weighted Bipartite Graph

When the server cluster is processing concurrent task requests, if the performance difference among servers is not fully considered, task allocation will be unreasonable, which will lead to an increase in task making span and a decrease in cluster resource utilization. As one of the core technologies of server cluster, load balancing is used to balance the load of each server by allocating tasks to each server through an algorithm before the task processing. Therefore, this paper proposes a dynamic load balancing algorithm based on optimal matching of weighted bipartite graph. First, we constructed a bipartite graph with servers and tasks as vertices. The management server collects the load indicators of each server in the cluster in real time, using the real-time processing rate of each server as the load indicator. Each edge of the bipartite graph is determined by comparing the expected completion time of the tasks with the load of each server. The degree of matching between each task amount and each server load capacity is defined as the weight matrix of the edges, and the bipartite graph is weighted to construct a weighted bipartite graph. The Kuhn-Munkres algorithm was used to solve the optimal matching of the weighted bipartite graph, and the optimal assignment of tasks to servers was achieved based on the result of the optimal matching. The proposed algorithm fully considers the differentiation of each task amount and each server load capacity. By building a server cluster example and conducting comparison experiments, it is demonstrated that the algorithm can achieve load balancing of the server cluster and improve the resource utilization efficiency of the cluster, while offsetting the extra time overhead caused by the algorithm.

A NOVEL DEEP LEARNING BASED METHOD FOR DETECTION AND COUNTING OF VEHICLES IN URBAN TRAFFIC SURVEILLANCE SYSTEMS

Abstract. In intelligent transportation systems (ITS), it is essential to obtain reliable statistics of the vehicular flow in order to create urban traffic management strategies. These systems have benefited from the increase in computational resources and the improvement of image processing methods, especially in object detection based on deep learning. This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP = 87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the proposed counting method was 94% with a real-time processing rate lower than 1.9.

Fast computation of Hahn moments on gray images using block representation

The moments constitute a well-known tool for image analysis and recognition tasks. The family of moments that has the most advantages is the discrete orthogonal moments. A set of these moments is the Hahn moments, as they have a great number of advantages in comparison with other sets of moments. The main disadvantage of moments, including Hahn moments, is the high computational cost, which is increased as higher order moments are involved in the computations, so the real-time analysis is hard to be done. We propose an effective approach for the computation of Hahn moments. The gray image is decomposed in a set of binary images that are named as bitplanes. The most significant bitplanes are represented using image block representation and their moments are computed fast using block techniques. The least significant binary images are substituted by a constant ideal image called “half-intensity” image, which has known Hahn moment values. The proposed method has low computational error, low computational complexity, and under certain conditions is able to achieve real-time processing rates.

A high throughput two-dimensional discrete cosine transform and MPEG4 motion estimation using vector coprocessor

In this work a configurable and scalable vector coprocessor for real time processing of MPEG4 motion estimation (ME) and two-dimensional DCT (2D DCT) is presented. A sequential DSP processor based on a reduced instruction set computer (RISC) processor architecture would require a frequency of 15 GHz for the real time processing of these two processes for a common intermediate format (CIF) sized sequence at 25 frames per second (fps). This frequency requirement will increase further if the image dimensions are increased. On the other hand our architecture on FPGA can achieve the real time processing rate at low frequency for CIF sized sequence and at higher frequency for full high definition (FHD) sequence for combined ME and 2D DCT. Due to configurable nature of the architecture and FPGA, this can be extended to higher dimensional image sequences. An important aspect of the architecture is that same datapath that is used for ME is also used for 2D DCT, with minor modification, leading to saving in area and time consumption. In addition the processor–coprocessor architecture has lower energy consumption and cost than the sequential processor.

Bioinspired point cloud representation: 3D object tracking

The problem of processing point cloud sequences is considered in this work. In particular, a system that represents and tracks objects in dynamic scenes acquired using low-cost sensors such as the Kinect is presented. An efficient neural network-based approach is proposed to represent and estimate the motion of 3D objects. This system addresses multiple computer vision tasks such as object segmentation, representation, motion analysis and tracking. The use of a neural network allows the unsupervised estimation of motion and the representation of objects in the scene. This proposal avoids the problem of finding corresponding features while tracking moving objects. A set of experiments are presented that demonstrate the validity of our method to track 3D objects. Moreover, an optimization strategy is applied to achieve real-time processing rates. Favorable results are presented demonstrating the capabilities of the GNG-based algorithm for this task. Some videos of the proposed system are available on the project website (http://www.dtic.ua.es/~sorts/3d_object_tracking/).

Embedded Hardware-Efficient Real-Time Classification With Cascade Support Vector Machines.

Cascade support vector machines (SVMs) are optimized to efficiently handle problems, where the majority of the data belong to one of the two classes, such as image object classification, and hence can provide speedups over monolithic (single) SVM classifiers. However, SVM classification is a computationally demanding task and existing hardware architectures for SVMs only consider monolithic classifiers. This paper proposes the acceleration of cascade SVMs through a hybrid processing hardware architecture optimized for the cascade SVM classification flow, accompanied by a method to reduce the required hardware resources for its implementation, and a method to improve the classification speed utilizing cascade information to further discard data samples. The proposed SVM cascade architecture is implemented on a Spartan-6 field-programmable gate array (FPGA) platform and evaluated for object detection on 800×600 (Super Video Graphics Array) resolution images. The proposed architecture, boosted by a neural network that processes cascade information, achieves a real-time processing rate of 40 frames/s for the benchmark face detection application. Furthermore, the hardware-reduction method results in the utilization of 25% less FPGA custom-logic resources and 20% peak power reduction compared with a baseline implementation.

Fast and Dense Depth Map Estimation for Stereovision Low-cost Systems

Depth map extraction or estimation is very important for many computer vision applications. However, this process involves high cost. Hence, it is interesting to find new approaches, which could yield faster implementations, when very accurate results are not necessary. These new implementations can be built up for the stereovision problem, taking advantage of the specific geometrical constraints of this kind of problem. The problem is solved by searching pixel-by-pixel matching in horizontal lines. In this article, we propose a novel approach to this problem, providing measurements of the processing time and accuracy achieved, as well as a qualitative comparison with other methods available in the literature. The results obtained could reach speeds around one order of magnitude faster than those proposed in other classical implementations. Thus, the proposed algorithm can run on cheap and low-performance hardware achieving real-time processing rates.

Multiplierless and fully pipelined JPEG compression soft IP targeting FPGAs

This paper presents the design of a soft IP for JPEG compression targeted for high performance in a FPGA device. The JPEG compressor architecture achieves high throughput with a deep and optimized pipeline and with a multiplierless datapath architecture. The JPEG compressor architecture was designed in a hierarchical and modular fashion and the details of the global architecture and of its modules are presented in this paper. A modular and strictly structural VHDL design is followed to develop the JPEG compressor soft IP. The VHDL codes were synthesized to Altera and Xilinx FPGAs. Synthesis results and relevant performance comparisons with related works are presented. Our high throughput compressor is able to compress 39.8 millions of pixels per second when mapped onto an Altera FLEX 10KE FPGA. Our JPEG soft IP mapped to FLEX 10KE low cost FPGA is able to compress 115 images per second in SDTV resolution (720 × 480 pixels). Considering this SDTV resolution our design is worthy as a core of an M-JPEG video compressor, reaching a real time processing rate of 30 fps, once mapped to the FLEX 10KE FPGA device.

Real-time gesture recognition system and application

In this paper, we consider a vision-based system that can interpret a user's gestures in real time to manipulate windows and objects within a graphical user interface. A hand segmentation procedure first extracts binary hand blob(s) from each frame of the acquired image sequence. Fourier descriptors are used to represent the shape of the hand blobs, and are input to radial-basis function (RBF) network(s) for pose classification. The pose likelihood vector from the RBF network output is used as input to the gesture recognizer, along with motion information. Gesture recognition performances using hidden Markov models (HMM) and recurrent neural networks (RNN) were investigated. Test results showed that the continuous HMM yielded the best performance with gesture recognition rates of 90.2%. Experiments with combining the continuous HMMs and RNNs revealed that a linear combination of the two classifiers improved the classification results to 91.9%. The gesture recognition system was deployed in a prototype user interface application, and users who tested it found the gestures intuitive and the application easy to use. Real time processing rates of up to 22 frames per second were obtained.

Motion Analysis on the Micro Grained Array Processor

Motion analysis plays a key role in video coding (e.g., video telephone, MPEG, HDTV) and computer vision systems (e.g., image segmentation, structure from motion). Motion estimation methods can be classified into three groups – matching-based, gradient-based, and frequency-based methods. The block matching algorithm (BMA) has been widely used for region matching in image coding, for example in MPEG (Motion Picture Expert's Group). Optical flow computation based on the spatio-temporal constraint equation has been broadly used in image segmentation to compute each pixel's velocity on a moving object. For both of these tasks, dedicated ASIC systems have been developed and widely used. Unfortunately, such systems have the disadvantage of restricted adaptability. The Micro Grained Array Processor (MGAP), which is a fine-grained, mesh-connected, SIMD array processor being developed at Penn State University, can provide a more regular, flexible, and efficient approach for solving, in real time, these two important computations.In this paper, we propose a new data flow scheme for an efficient, systolic, full-search BMA on programmable array processors so that we can process as many adjacent template blocks as possible in unison in order to reduce the data memory accesses. In particular we present an efficient implementation of the BMA on the MGAP. As a result, the BMA for the MPEG SIF video format (352 × 240 pixels) with a block size of 16 × 16 pixels, a displacement range of 16 pixels, and frame rate of 30 frames/sec can be computed at real-time processing rates on the MGAP. We also show a real-time mapping to the MGAP of the optical flow computation for images of size 256 × 256 pixels.