Number Of Consecutive Frames Research Articles

DRIVER DROWSINESS DETECTION SYSTEM

The proposed system utilizes facial landmark detection and eye aspect ratio calculation to detect drowsiness in real-time from a webcam feed. It employs the dlib library for face detection and facial landmark prediction. The eye aspect ratio (EAR) is computed from the distances between specific points on the eyes. When the EAR falls below a predefined threshold for a certain number of consecutive frames, it triggers an alert indicating potential drowsiness. The alert includes visual cues on the video feed, an auditory alert using the playsound library, and even sends an instant WhatsApp message using pywhatkit to notify a designated contact about the drowsiness. This system serves as a useful tool for monitoring driver fatigue or alertness in various contexts, promoting safety and awareness. Its capacity to swiftly identify and alert users to the onset of drowsiness represents a significant leap forward in safety technology. By mitigating the risk of accidents stemming from fatigue-induced impairments, this project underscores the pivotal role of computer vision in real-time human behavior monitoring, offering a versatile and invaluable tool for fostering safety and awareness across various domains. Keywords: Drowsiness Detection; Python; Whatsapp messaging; Webcam feed; Computer Vision; Alert System; Real-time monitoring; Eye-Aspect Ratio; Frame Processing; Facial Recognition.

Multi-Frame Track-Before-Detect Algorithm for Maneuvering Target Tracking

Multi-frame track-before-detect (MF-TBD) is a model-based batch processing method. Assuming a particular model for the evolution of target states (e.g. a constant velocity model) within a batch processing time, MF-TBD integrates the target energy by taking advantage of the space-time correlations between a number of consecutive frames. Its performance is known to be heavily dependent on the accuracy of the motion model, and to substantially degrade when target maneuvers occur and motions do not follow the presumed model. We make two contributions towards addressing this problem. Firstly, we analyze and summarize the direct strategies to incorporate the effect of target maneuvers into MF-TBD. Our analysis shows that although these strategies are straightforward to implement, they either suffer from considerable performance loss or are computationally expensive. Motivated by the analysis, we propose a general measurement-directed (MD) strategy to address target maneuvers. It carries out an on-line study of target dynamics from the observations, and is capable of achieving both low computational complexity and high adaptability to different target maneuvers. Secondly, as the proposed MD strategy is a general framework without any particular model assumptions, we further derive its detailed implementation equations for linear motion and measurement models. Simulation results for various tracking scenarios are presented to demonstrate the effectiveness of the proposed MD strategy.

TiVGAN: Text to Image to Video Generation With Step-by-Step Evolutionary Generator

Advances in technology have led to the development of methods that can create desired visual multimedia. In particular, image generation using deep learning has been extensively studied across diverse fields. In comparison, video generation, especially on conditional inputs, remains a challenging and less explored area. To narrow this gap, we aim to train our model to produce a video corresponding to a given text description. We propose a novel training framework, Text-to-Image-to-Video Generative Adversarial Network (TiVGAN), which evolves frame-by-frame and finally produces a full-length video. In the first phase, we focus on creating a high-quality single video frame while learning the relationship between the text and an image. As the steps proceed, our model is trained gradually on more number of consecutive frames. This step-by-step learning process helps stabilize the training and enables the creation of high-resolution video based on conditional text descriptions. Qualitative and quantitative experimental results on various datasets demonstrate the effectiveness of the proposed method.

Quantitative Doppler in musculoskeletal ultrasonography - suboptimal performance of both experienced and in-training sonographers in selection of the highest Doppler signal image from cine-loops.

Doppler ultrasonography assessment is mandatory nowadays for the complete description of rheumatic disease activity.Initially it was performed in semi quantitative way but recently the (fully) quantitative assessment is gaining more interest. In quantitative assessment, the ratio between total colorized and total pixels (CTR) is computed for the whole image or just for theregion of interest (ROI). The frame with the highest amount of Doppler signal (also called worst case scenario image - WCSI) isusually the only one analyzed. The technique requires a very precise identification of WCSI from a certain number of consecutiveframes, captured from the same position of the US probe, (and in most cases this is done manually). Our study examined theability of both experienced and in-training sonographers to identify WCSI using a computerized analytical system as the goldstandard. The study analyzed 480 frame selections done in two distinct exercises. The WCSI and other 3images with a 5%, 10% and respectively 20% lower level of CTR compared with WCSI were packed in one selection. All framesemerging from the same video clip were randomly presented to six experienced and six in training sonographers; the request wasto select the frame with the highest CTR (WCSI) from each package (twenty packages in total). A similar exercise was performedwith CTRs decreasing in steps of 2%. In the first exercise the WCSI was correctly identified in 79.1% cases and in67% of cases in the 2nd exercise. The interobserver agreement between experienced and in-trainer evaluators for the 1st exercisewas 0.78 and 0.4 in the 2nd exercise. Using computerized analysis as the gold standard, we demonstrated a largeheterogeneity across sonographers regarding their ability to identify the best Doppler image even from a small group of frames.

CAMHID: Camera Motion Histogram Descriptor and Its Application to Cinematographic Shot Classification

In this paper, we propose a nonparametric camera motion descriptor for video shot classification. In the proposed method, a motion vector field (MVF) is constructed for each consecutive video frame by computing the motion vector (MV) of each macroblock. Then, the MVFs are divided into a number of local region of equal size. Next, the inconsistent/noisy MVs of each local region are eliminated by a motion consistency analysis. The remaining MVs of each local region from a number of consecutive frames are further collected for a compact representation. Initially, a matrix is formed using the MVs. Then, the matrix is decomposed using a singular value decomposition technique to represent the dominant motion. Finally, the angle of the most variance retaining principal component is computed and quantized to represent the motion of a local region by using a histogram. In order to represent the global camera motion, the local histograms are combined. The effectiveness of the proposed motion descriptor for video shot classification is tested by using a support vector machine. First, the proposed camera motion descriptors for video shots classification are computed on a video data set consisting of regular camera motion patterns (e.g., pan, zoom, tilt, static). Then, we apply the camera motion descriptors with an extended set of features to the classification of cinematographic shots. The experimental results show that the proposed shot level camera motion descriptor has a strong discriminative capability to classify different camera motion patterns of different videos effectively. We also show that our approach outperforms state-of-the-art methods.

An efficient fire detection method based on orientation feature

This paper proposes a novel method for reliable fire detection. The burning fire usually causes rich moving features in terms of directions, which can offer the best chance to distinguish between the fire region and the non-fire one. Motivated by this observation, we design a novel orientation feature to represent this characteristic. Based on this feature, a method is proposed to detect the fire efficiently. First, fire color is utilized to extract the fire candidate areas from the surveillance video. Then, the direction is obtained by computing the optical flow for each pixel in the candidate area. The directions are discretized to four parts. By counting the percentage of pixels whose moving directions fall into these four parts in a period of time, and combining with the two parameters, i.e., both of the number of frames without the moving directions and the number of consecutive frames in the candidate area, we use these six parameters as the fire orientation feature. In the end, by training a support vector machine (SVM) classifier with the input of our fire orientation feature, the candidate area is judged whether it is a fire. Our main contribution is that we design the novel fire orientation feature. The feature can not only characterize the fire intrinsic dynamic properties accurately but also is very efficient. Compared with the art-of-state methods, the experimental results confirm that our approach significantly improves the accuracy of fire detection and impressively decreases the false alarm rate. The detection speed of our approach is also very competitive with the art-of-state fire detection methods.

MMSE-Based Missing-Feature Reconstruction With Temporal Modeling for Robust Speech Recognition

This paper addresses the problem of feature compensation in the log-spectral domain by using the missing-data (MD) approach to noise robust speech recognition, that is, the log-spectral features can be either almost unaffected by noise or completely masked by it. First, a general MD framework based on minimum mean square error (MMSE) estimation is introduced which exploits the correlation across frequency bands to reconstruct the missing features. This framework allows the derivation of different MD imputation approaches and, in particular, a novel technique taking advantage of truncated Gaussian distributions is presented. While the proposed technique provides excellent results at high and medium signal-to-noise ratios (SNRs), its performance diminishes at low SNRs where very few reliable features are available. The reconstruction technique is therefore extended to exploit temporal constraints using two different approaches. In the first approach, time-frequency patches of speech containing a number of consecutive frames are modeled using a Gaussian mixture model (GMM). In the second one, the sequential structure of speech is alternatively modeled by a hidden Markov model (HMM). The proposed techniques are evaluated on Aurora-2 and Aurora-4 databases using both oracle and estimated masks. In both cases, the proposed techniques outperform the recognition performance obtained by the baseline system and other related techniques. Also, the introduction of a temporal modeling turns out to be very effective in reconstructing spectra at low SNRs. In particular, HMMs show the highest capability of accounting for time correlations and, therefore, achieve the best results.

Scalable and energy-efficient packet switches based on multi-granular forwarding operations

Telecommunication networks are evolving due to rapid growth of internet traffic and the necessity to satisfy the new requirements of emerging packet services. Upgrading of network devices so as they can allow scalable and full line rate traffic aggregation and eliminate any internal performance bottlenecks is crucial. In packet switching devices, since data plane functionalities need to be executed for each incoming packet and power per given bandwidth is strongly related to the amount of processing on this data, packet processing at ultra-high rate is becoming the major challenge. The main trend to address such power consumption and scalability issues is to bypass packet switches by switching traffic at lower layers. This results in the packet optical transport network approach, where packet switching provides flexible end-to-end connectivity based on tunnel encapsulation while wavelength switching, exploiting optical bypass, allows reducing electrical switch size at transit nodes. There are also opportunities consisting in simplifying packet switch functionalities or designing completely different packet switch architectures. In this paper a new Ethernet aggregation and switching solution with potentialities to simplify and scale Ethernet switch forwarding functionality is proposed. This solution, based on a burst-basis transmission compliant with the Ethernet Standard, is able to maintain flexibility and any to any connectivity deriving from the connectionless nature of Ethernet. At the same time, it provides Ethernet technology with efficient aggregation capabilities allowing to reduce processing of transit traffic. This is allowed thanks to an Ethernet burst structure conceived as a variable number of consecutive frames of the same connection preceded by a proprietary burst control frame carrying information necessary for burst data frames classification. As a result, burst control frames experience the conventional Ethernet switch packet processing while data frames are mapped on the corresponding queue/output port according to the result of control frames classification. The proposed solution also provides the possibility to recognize data frames of the same burst through a proprietary inter-frame gap inserted among them; that allows to dynamically adapt burst size to the available bandwidth at transit nodes in order to limit frame delay and jitter and to support intermediate grooming. These features make it a very competitive approach in the context of packet optical transport being able to support dynamic multi-granular switching. The proposed solution has been validated by estimating its efficiency in terms of energy consumption with respect to a commercial packet switch. The impact of burst transmission on packet delay and jitter across ring and mesh networks has been also evaluated through different sets of simulations.

Video post processing: low-latency spatiotemporal approach for detection and removal of rain

In this study, a novel, efficient and simple algorithm for detection and removal of rain from video using spatiotemporal properties is proposed. Advantageously, the spatiotemporal properties are involved to separate rain pixels from non-rain pixels. It is thus possible by way of the proposed algorithm to involve less number of consecutive frames, reducing the buffer size and delay. It works only on the intensity plane which further reduces the complexity and execution time significantly. This new algorithm does not assume the shape, size and velocity of raindrops which makes it robust to different rain conditions. Proposed method reduces the buffer size, which reduces the system cost, delay and power consumption. For performance evaluation, in addition to miss & false detection a new metric spatiotemporal variance is introduced. Results show that the proposed algorithm outperforms the other rain removal algorithms.

Laptop-sized Echocardiography Machine Versus Full-sized Top-of-the-Line Machine: A Comparative Study

Echocardiography has become an essential part of the evaluation of patients in a wide variety of clinical settings 1 and an invaluable tool in the treatment of patients who are critically ill in the intensive care unit (ICU) and emergency department. Echocardiography can be performed at the bedside, thus, avoiding the need to transport patients who are acutely ill. However, there still remain obstacles to using portable echocardiography. The current commonly available echocardiography machines are bulky and heavy (400 lb). Bringing the echocardiography machine to the bedside is time-consuming and may cause damage to valuable equipment and personnel in that process. Often the images obtained are suboptimal secondary to faulty positioning of the machine as a result of space constraints, difficult positioning of patients, and poor lighting conditions. There have been prior attempts to design a small, truly portable echocardiography machine. Several products have been previously manufactured. The sizes ranged from laptop-sized devices to even smaller units (hand-carried devices) of less than 6 lb in weight. The earliest models had limited functionality; however, newer models have nearly all the standard features present on full-sized echocardiography equipment. Nonetheless, the performance of these units has proved to be inferior to that of the standard, bulky echocardiographic machines. 2-8 Several studies were done to evaluate the diagnostic quality of these hand-carried ultrasound systems. Although information obtained with these studies helped patient care and treatment, the sensitivity of these units to accurately identify different cardiac pathologies, compared with standard echocardiography machines, was lower (70%-80%). Thus, the use of these hand-carried ultrasound machines seemed promising and appropriate (in adequately trained hands) for limited studies with a clear, predefined clinical question. 9-12 Recently a new, laptop-sized echocardiography machine has been introduced into the market (Vivid I, GE, Milwaukee, Wis). This machine is the size of an average laptop computer and weighs approximately 10 lb. It is easily transported and can be used comfortably in every department in the hospital. It has been approved by the Food and Drug Administration for clinical use. The machine has all the modalities of a standard echocardiography machine: M-mode, 2-dimensional imaging, spectral Doppler, continuous wave and pulsed wave Doppler, harmonic imaging, and tissue Doppler. There is also a prototype that carries unique software for speckle tracking, non-Doppler 2-dimenstional strain analysis. Automated speckle tracking is a new means by which endocardial border and ventricular myocardium can be detected. It analyzes motion by tracking speckles (natural acoustic markers) in the ultrasonic image in two dimensions. These markers are statistically equally distributed throughout the myocardium. The size of these elements is 20 to 40 pixels. Each speckle can be identified and followed accurately over a number of consecutive frames. These markers, within the ultrasonic image, are tracked by calculating frame-to-frame changes using a sum of absolute difference algorithm. Novel software is now available that allows automatic speckle tracking with subsequent calculation of ejection fraction (EF), tissue velocities, strain, and strain

Keratocyte Density

The ConfoScan 3 clinical confocal microscope provides sharper images of cells in the stroma than the Tandem Scanning confocal microscope does. In this study, we compared volumetric densities of stromal cells determined from images recorded by these two instruments. Fifty corneas of 25 normal subjects were examined by confocal microscopy, first by using a Tandem Scanning confocal microscope and then by using a ConfoScan 3 confocal microscope. Bright objects, assumed to represent keratocytes, were counted in a known area of two frames selected from the mid-stroma. The effective depth of the sample volume represented by each frame was estimated from the number of consecutive frames and the corresponding distance that selected cells were visible and countable during a scan. Density was the number of visible cells in the sample area divided by the effective sample volume. The effective focal depth of the Tandem Scanning microscope was 11.9 +/- 2.6 microm (mean +/- SD), and was 25.9 +/- 7.1 microm for the ConfoScan 3. Mean cell density at mid stroma was 23,013 +/- 4,420 cells/mm(3) with the Tandem Scanning microscope and 23,996 +/- 2,898 cells/mm(3) with the ConfoScan 3. This difference was not significant (P = 0.15). In normal corneas, the ConfoScan 3 and the Tandem Scanning confocal microscopes indicate stromal cell densities that are not significantly different from each other. Estimates of cell density from both instruments require an accurate estimate of the effective depth of the sample volume; this depth is approximately 2.2 times greater with the ConfoScan 3. This difference must be considered when comparing results from studies that use one instrument with results from studies that use the other.

Speech processing apparatus and method

An apparatus is provided for checking the consistency between two training words which can be used in, for example, a speech recognition or verification system. Two training examples are aligned using a dynamic programming alignment process and an average frame score is calculated from the alignment results together with the worst score in a number of consecutive frames. These values are then compared with similar values obtained from training examples which are known to be consistent to determine if the training examples are consistent.

Variable temporal-length 3-D discrete cosine transform coding

Three-dimensional discrete cosine transform (3-D DCT) coding has the advantage of reducing the interframe redundancy among a number of consecutive frames, while the motion compensation technique can only reduce the redundancy of at most two frames. However, the performance of the 3-D DCT coding will be degraded for complex scenes with a greater amount of motion. This paper presents a 3-D DCT coding with a variable temporal length that is determined by the scene change detector. Our idea is to let the motion activity in each block be very low, while the efficiency of the 3-D DCT coding could be increased. Experimental results show that this technique is indeed very efficient. The present approach has substantial improvement over the conventional fixed-length 3-D DCT coding and is also better than that of the Moving Picture Expert Group (MPEG) coding.