Search

Visual impairment/color-blindness (VICB) can be challenging in many situations especially while crossing a pedestrian/cross-walk or driving a vehicle. A Traffic Light Recognition System (TLRs) may help in the accurate detection of traffic lights using a hand-held mobile device. TLR helps in reduction of accidental mortality rates. It will also help improving transportation and mobility for old aged and differently-abled people. TLR system detects the presence of traffic light in the environment and incorporates guiding assistance by notifying its color and shape to a VICB person. There may be enormous challenges in correct detection of Traffic Lights, involving non-working lights, illuminations, ego-vehicles, weather, trees, and other obstructions. A detailed discussion of the TLR System is provided for data acquisition, pre-processing, localization,feature extraction and verification stages. Finally, the conclusions are drawn and possible future scope of the field is discussed.

The current development of technology provides us with access to increasingly complex computers with enormous computing power. Reducing their dimensions and increasing availability on the market resulted in a reduction in their manufacturing costs. Occurred situation causes increased development of applications based on artificial intelligence methods. On the other hand, the user uses a web browser to access the computer with computing power much greater than his desktop computer. Such possibilities are offered to us by the most popular platform offering free computing power and the ability to store data which is Google Collaboratory. Greater computing power significantly speeds up the complex calculations required to train artificial neural networks, which are becoming more and more popular.

In this paper, we study the iterative self-supervised pretraining procedure for the Tatar language speech recognition system. The complete recipe includes the use of base pre-trained model (the multilingual XLSR model or the Librispeech (English) Wav2Vec 2.0 Base model), the next step was a “source” self-supervised pre-training on collected Tatar unlabeled data (mostly broadcast audio), then the resulting model was used for additional “target” self-supervised pretraining on the annotated corpus (target domain, without using labels), and the final step was to fine-tune the model on the annotated corpus with labels. To conduct the experiments we prepared a 328-h unlabeled and a 129-h annotated audio corpora. Experiments on three datasets (two proprietary and publicly available Common Voice as the third one) showed that the first “source” pretraining step allows ASR models to show on average 24.3% lower WER, and both source and target pretraining - 33.3% lower WER than a simple finetunes base model. The resulting accuracy for the Common Voice (read speech) test dataset is WER 5.37%, on the private TatarCorpus (read clean speech) is 4.65%, and for the spontaneous speech dataset collected from the TV shows is 22.6%, all of the results are the best-published results on these datasets. Additionally, we show that using a multilingual base model can be beneficial for the case of fine-tuning (10.5% less WER for this case), but applying self-supervised pretraining steps eliminates this difference.

Engagement recognition is a growing domain in paralinguistics evaluation due to its importance in many human-computer and human-robot applications. Current requirements for such systems are not only to interact, but also to engage the user into interaction as long as possible. To do so, the machine should differ among different levels of engagement to adjust its behavior properly. However, actual models are still far from it, partially due to data quality – usually, engagement recognition datasets are highly biased towards the high-engagement levels, because they are more often and naturally expressed by humans during interaction context. Thus, currently, the development of a reliable engagement recognition system able to detect all engagement levels is necessary. To facilitate it, we introduce a deep learning engagement recognition framework in the context of the DAiSEE corpus, which is a highly imbalanced dataset. We showed that the metric used formerly for evaluating the performance of the models on the DAiSEE dataset is inadequate due to its imbalance and conducted extensive experiments on DAiSEE, suggesting a new baseline performance based on the Unweighted Average Recall metric.

This paper proposes relation-based sequence representation that enhances offline handwritten mathematical expressions (HMEs) recognition. Commonly, a LaTeX-based sequence represents the 2D structure of an HME as a 1D sequence. Consequently, the LaTeX-based sequence becomes longer, and HME recognition systems have difficulty in extracting its 2D structure. We propose a new representation for HMEs according to the relations of symbols, which shortens the LaTeX-based representation. We use an offline end-to-end HME recognition system that adopts weakly supervised learning to evaluate the proposed representation. Recognition experiments indicate that the proposed relation-based representation helps the HME recognition system achieve higher performance than the LaTeX-based representation. In fact, the HME recognition system achieves recognition rates of 53.35%, 52.14%, and 53.13% on the dataset of the Competition on Recognition of Online Handwritten Mathematical Expressions (CROHME) 2014, 2016, and 2019, respectively. These results are more than 2 percentage points higher than the LaTeX-based system.

Optical Character Recognition (OCR) plays an essential role in nowadays life, which contributes to solving problems in terms of timing and accuracy of documents. The use of OCR in the health sector can help solve problems of drug handling or inventorying in drug banks to prevent unnecessary risks. However, if you apply existing OCR meth- ods such as Tesseract or EasyOCR, it will be challenging to find out the name of the medicine or the medicine ingredient in a prescription. In this paper, we propose a system to help find the medicine names from the prescription image. We then provide users with information on the identified medicine names. Methods are built by combining and transforming many existing identity models. In addition, we have successfully developed an application running on the Android platform to get feedback on improving the system and want to help them get more information about the drugs they are using. Experimental results show that the model recognizes drug names quite well on a given database, even with medium resolution photos.

Unmanned Aerial Vehicles (UAV) have the potential to improve the first recognition activities that are necessary for humanitarian relief operations in the aftermath of a sudden disaster. Rapid mapping of the affected area is one example of these activities. UAVs provide solutions to the challenge of obtaining reliable information about the status and location of an affected zone and the coordination of the different rescue teams involved in relief operations. This paper proposes a UAV-based rapid mapping system for the first stage of recognition in a post-disaster situation, helping to recognize the status or level of damage at a specific zone. This recognition system is modelled based on an adapted integrated capacitated location-routing (CLRP) model corresponding to the operating specifications of UAVs that represent challenging constraints. This model seeks to determine the optimal location of the UAV’s hubs that should be placed before the disaster happens, and the optimal routing of UAVs after the disaster has happened. The adapted model is tested with data generated trying to emulate a small city. The results show a near-optimal location, number of UAV needed, and the routes for each one. From these results, organizational insights are provided to put in place the system. Also, relevant research directions in UAV location routing for rapid mapping are proposed.

This paper presents a dialect recognition system for the Kurdish language using speaker embeddings. Two main goals are followed in this research: first, we investigate the availability of dialect information in speaker embeddings, then this information is used for spoken dialect recognition in the Kurdish language. Second, we introduce a public dataset for Kurdish spoken dialect recognition named Zar. The Zar dataset comprises 16,385 utterances in 49h-36min for five dialects of the Kurdish language (Northern Kurdish, Central Kurdish, Southern Kurdish, Hawrami, and Zazaki). The dialect recognition is done with x-vector speaker embedding which is trained for speaker recognition using Vox-celeb1 and Voxceleb2 datasets. After that, the extracted x-vectors are used to train support vector machine (SVM) and decision tree classifiers for dialect recognition. The results are compared with an i-vector system that is trained specifically for Kurdish spoken dialect recognition. In both systems (i-vector and x-vector), the SVM classifier with 86% of precision results in better performance. Our results show that the information preserved in the speaker embeddings can be used for automatic dialect recognition.

Digit recognition has a vital use in multiple human-machine interaction applications. It is used in telephone-based services, such as dialing systems, airline reservation systems, different bank transactions, and price extraction. This research aims to develop a new Convolution Neural Network (CNN) based spoken digits recognition system for the Arabic digits. The developed system used a classification approach to perform the recognition task. First, the Mel frequency cepstral coefficients of the spoken digits were conducted and reduced in the convolution phase. Then in the classification phase, the most appropriate digit label for the testing utterances is produced. The proposed approach has shown a remarkable performance when compared to similar systems. The recognition system achieved a 99% correct digit recognition compared to 98% using Recurrent Neural Networks based digit recognition system.

The aim of the work is to develop a safety subsystem that is part of the control system of a robotic laser welding complex. The main tasks to be solved: the development of hardware implementation of the complex security circuit and the development of security algorithms for the main technological equipment of the complex. To solve the tasks, an analysis of the main technological equipment of the complex and laser welding technology was made. A list of processed contingencies was compiled. The specifications of the local security systems built into the main equipment were analyzed. Local security facilities are integrated into a common security circuit of the complex. Algorithms for the operation of the safety subsystem elements of the robotic laser welding complex were developed. In the analysis of existing means of protection against laser radiation, it was concluded that it is necessary to develop an external protective fence of the complex with an active recognition system for the beam entering the fence. The main results of the work are the scheme for connecting local security systems of the main technological equipment into a single complex security circuit and the algorithms of the complex’s security subsystem.

We present a vision-based activity recognition system for centrally connected humanoid robots. The robots interact with several human participants who have varying behavioral styles and inter-activity-variability. A cloud server provides and updates the recognition model in all robots. The server continuously fetches the new activity videos recorded by the robots. It also fetches corresponding results and ground-truths provided by the human interacting with the robot. A decision on when to retrain the recognition model is made by an evolving performance-based logic. In the current article, we present the aforementioned adaptive recognition system with special emphasis on the partitioning logic employed for the division of new videos in training, cross-validation, and test groups of the next retraining instance. The distinct operating logic is based on class-wise recognition inaccuracies of the existing model. We compare this approach to a probabilistic partitioning approach in which the videos are partitioned with no performance considerations.

The system of radar signals detection, analysis and recognition was described. The modern electronic recognition systems should react fast and with great accuracy in the extremely complex electromagnetic environment. The binary decision tree was applied at the beginning of grouping signals received from unknown sources. The paper presents some clustering methods to radar signal recognition based on the mathematical criteria. The concepts of this technique are described. The experiment results obtained for nearest neighbour method are presented. Clustering algorithm was tested for different methods of objects grouping and for various distance measures.

The texture classification problem is widely applied for ground-based cloud images recognition due to its efficiency. Local Binary Pattern and its variants are usually investigated to represent cloud images. The appropriate choice of color space might enhance performance of the recognition system for many applications. In this paper, we propose a comparative study to select the best candidate’s space for coding cloud images by fusing multiple texture description methods. The proposed approach is evaluated on the SWIMCAT database.

In unmanned stores, automated checkout is an integral part of the process, and the checkout is usually completed by expensive identification machines. Some unmanned stores lacking banknotes and coins only provide credit cards, EasyCard, or QR code payment methods, sometimes that cause the difficulty of payment when they check out. This research is aimed at the coin recognition for images. It processes the images using OpenCV, and substitutes into the trained convolutional neural network (CNN) for identification. The result of the research shows that the accuracy of the model identification is 94%, and it can be used to identify more than one coin.

Plant recognition is an important problem and can be performed manually by specialists, but in this case, a lot of time is required and therefore is low-efficiency. Thus, automatic plant recognition is an important area of research. In this paper, we propose an ensemble of models to increase the performance of plant recognition. The ensemble of models presents a composite model which has two-level architecture. At the first level of the stacking model, convolutional neural networks are used, which demonstrate high performance in solving problems of object recognition. At the second level, gradient boosting methods are used. The model is taught using an open database of plant images containing 12 different species. Experiments with a plant dataset show that the proposed model is significantly better than other classification methods. High classification accuracy makes the model very useful for supporting the plant recognition system for working in real conditions.

Deaf and hearing-impaired individuals who communicate using sign language face several communication difficulties. Because the vast majority of people do not know sign language, the need for a sign language translator is growing significantly, especially for Arabic sign language (ArSL). Today, technology plays a significant role in people’s lives. Leap Motion technology can be used to address the aforementioned issues and improve communication between Saudi Arabia’s deaf and hearing individuals. In this study, we investigated the possibility of using a Leap Motion system to provide continuous ArSL recognition for two-way communication to improve communication between deaf and hearing individuals in terms of speed and independence. The system translates ArSL into spoken words for hearing individuals and transcribes spoken Arabic language into text for deaf individuals.

Food image recognition has been extensively investigated during the last decade and had multiple useful applications for monitoring food calories and analyzing people’s eating habits to ensure better health. In this paper, we study a Vietnamese food recognition system using Convolutional Neural Networks (CNNs) based features. We manually collect one dataset for Vietnamese food classification with 13 categories and 8903 images. For learning a proper food classifier, we conduct brief analytics by comparing hand-crafted features and CNNs based features (including AlexNet, GoogleNet, ResNet50, ResNet101v2, and InceptionResnetv2) and choosing top K accuracy for measuring the performance of each model. The experimental results show that InceptionResnetv2 can achieve the best performance among all these techniques. We aim at publishing our codes and datasets for giving and additional contribution to the research community related to the Vietnamese food recognition problem.

Many drink packages have expiry dates written in dot matrix characters (digits and non-digits, e.g., slashes or dots). We collected images of these packages and trained two existing deep neural networks (DNNs) to combine and form a system for detecting and recognizing expiry dates on drink packages. One of the DNNs is an object-detection DNN and the other is a character-recognition DNN. The object-detection DNN alone can localize the characters written on a drink package but its recognition accuracy is not sufficient. The character-recognition DNN alone cannot localize characters but has good recognition accuracy. Because the system is a combination of these two DNNs, it improves the recognition accuracy. The object-detection DNN is first used to detect and recognize the expiry date by localizing and obtaining the size of the character. It then scans the expiry-date region and clips the image. The character-recognition DNN then recognizes the characters from the clipped images. Finally, the system uses both DNNs to obtain the most accurate recognition result based on the spacing of the digits. We conducted an experiment to recognize the expiry dates written on the drink package. The experimental results indicate that the recognition accuracy of the object-detection DNN alone was 90%, that of the character-recognition DNN alone was also 90%, and that combining the results of both DNNs was 97%.

Automatic License Plate Recognition (ALPR) systems are used in many real-world applications, such as road traffic monitoring and traffic law enforcement, and the use of deep learning can result in efficient methods. In this work, we present an ALPR system efficient for edge computing, using a combination of MobileNet-SSD for vehicle detection, Tiny YOLOv3 for license plate detection and OCR-net for character recognition. This method was evaluated in two datasets on a NVIDIA Jetson TX2 system, obtaining 96.87% of accuracy and 8 FPS of framerate in a public real-world scenario dataset and achieving 90.56% of accuracy and 11 FPS of framerate in a private dataset of traffic monitoring images, considering the recognition of at least six characters. It is faster than related works with similar deep learning approaches, that achieved at most 2 FPS, and slightly inferior in accuracy, with less than 10% of difference in the worst scenario. This shows the proposed method is well balanced between accuracy and speed, thus, suitable for embedded devices.

Solid waste management is an essential task to be carried out in day-to-day life. So an automated recognition system using deep learning algorithm has been implemented to classify wastes as biodegradable and non-biodegradable. Efficient segregation of solid wastes helps to reduce the amount of waste buried in the ground, thereby improving the recycling rate, and safeguards the soil from pollution.

As fingerprint recognition systems have become increasingly adopted within a range of technology applications over the last decade, so too has their attention within emerging research. However, although this increased attention has led to an enhancement of the software and algorithms behind this recognition process, the majority of research has still not addressed the issues of incorrect rotation or proximity between the finger and the device. Current systems assume that the direction of the imprinted finger will align with that of the target fingerprint image; this decreases the accuracy of fingerprint recognition across a variety of finger orientations and scenarios.

This work presents an application of wearable technology and machine learning techniques for automatic identification of the use time of hand-held vibrating tools in the workplace. The proposed system is an automatic recognition system based in a commercial smartwatch that can be used in tasks related to the risk assessment produced by exposure to vibrations that affects the hand-arm system. The system can identify with high accuracy, three types of machine families and identify a single model within a single tool family. At present, it is possible to use intelligent wearable devices for the development of technological solutions that can help to improve the current methodologies for quantifying the effects produced by the exposure to hand-held vibrating tools, as well as its level of impact on workers’ health. In the near future, the use of systems similar to this may allow the analysis of the occupational risks produced by exposure to mechanical vibrations in the workplace in an automated, precise and low-cost way, as well as being part of risk management systems integrated into the concept of industry 4.0.

Automatic target detection and recognition is the cornerstone of the intelligent unmanned systems to realize higher-level tasks. In this paper, the deep learning algorithm of Faster R-CNN was studied in depth, and the target detection model is designed combining the RPN network and the fast R-CNN. The target detection and recognition device with the ability of image acquisition and intelligent processing was also designed. Combining the device with the Faster R-CNN model, the automatic target detection and recognition system was developed. At last, the VGG-16 model was adopted for training the detection model, and the system was used for target detection experiments. The results show that the recognition accuracies of the system for the visible light images of trucks and tanks are 89.7% and 90.3%, respectively, and that for infrared images of tanks is 63.7%. Therefore, a good recognition effect has been achieved. This work provides a reference for the application of deep learning algorithms in the field of automatic target detection and recognition.

The field of biometrics is currently an area that is both very interesting as well as rapidly growing. Among various types of behavioral biometrics, human gait recognition is worthy of particular attention. Unfortunately, one issue which is frequently overlooked in subject-related literature is the problem of the changing quality of a biometric system in relation to tests that are repeated after some time. The present article describes tests meant to assess the accuracy of a human gait recognition system based on Ground Reaction Forces in time removed repeated trials. Both the initial testing as well as the repeated trials were performed with the participation of the same 40 people (16 women and 24 men) which allowed the recording of nearly 1,600 stride sequences (approximately 800 in each trial). Depending on the adopted scenario correct recognition ranged from 90.4% to 100% of cases. These results indicate that the biometric system had greater problems with recognition the longer the period of time which passed since the first trials. The present article also analyzed the impact of footwear change in the second series of testing on recognition results.

The human recognition is achieved easier and cheaper and the single modality employed for the recognition faces a lot of challenges due to the environmental factors. This paper proposes a multimodal recognition system based on the Multi-Support Vector Neural Network (Multi-SVNN). The algorithm proposed is Glowworm Penguin Search Optimization Algorithm (GwPeSOA), which is the modification of the Glowworm Optimization Algorithm (GOA) with the Penguin Search Optimization Algorithm (PeSOA). The proposed method employs ear and the palm vein modalities and the features of the ear image is obtained using the proposed BiComp masking method of feature extraction, whereas the features from the palm vein is extracted using the Local Binary Pattern method. The features obtained are applied to the Multi-SVNN classifier to recognize with good accuracy and the proposed BiComp Mask offers the robust features for the extraction. The experimentation using the proposed method attained a better accuracy, specificity, and sensitivity.

For any recognition system, the ability to identify novel class samples during inference is an important aspect of the system’s robustness. This problem of detecting novel class samples during inference is commonly referred to as Multiple Class Novelty Detection. In this paper, we propose a novel method that makes deep convolutional neural networks robust to novel classes. Specifically, during training one branch performs traditional classification (referred to as global inference), and the other branch provides patch-level information to keep track of the class-specific activation patterns (referred to as local inference). Both global and local branch information are combined to train a novelty detection network, which is used during inference to identify novel classes. We evaluate the proposed method on four datasets (Caltech256, CUB-200, Stanford Dogs and FounderType-200) and show that the proposed method is able to identify novel class samples better compared to the other deep convolutional neural network-based methods.

In this paper, we investigate the benefit of 3D hand skeletal information to the task of sign language (SL) recognition from RGB videos, within a state-of-the-art, multiple-stream, deep-learning recognition system. As most SL datasets are available in traditional RGB-only video lacking depth information, we propose to infer 3D coordinates of the hand joints from RGB data via a powerful architecture that has been primarily introduced in the literature for the task of 3D human pose estimation. We then fuse these estimates with additional SL informative streams, namely 2D skeletal data, as well as convolutional neural network-based hand- and mouth-region representations, and employ an attention-based encoder-decoder for recognition. We evaluate our proposed approach on a corpus of isolated signs of Greek SL and a dataset of continuous finger-spelling in American SL, reporting significant gains by the inclusion of 3D hand pose information, while also outperforming the state-of-the-art on both databases. Further, we evaluate the 3D hand pose estimation technique as standalone.

The mobility of people who have suffered a degenerative disease or an accident is partially or totally reduced, which limits their locomotive independence. Therefore, this paper presents a proposal that facilitates the mobility of people suffering from moderate levels of amyotrophic lateral sclerosis (ALS). A control system has been adapted to an electric wheelchair to provide it with a certain degree of intelligence. The acquisition of multimedia data is done with a small camera adapted to a glasses frame that the person must use. For eye patterns tracking, a recognition system is performed using the LabVIEW software environment. The control system that regulates the movement of the wheelchair was designed on the Raspberry Pi embedded board as a low-cost proposal. Experimental tests and user surveys validate the correct operation of this device.

In this paper, we present a system based on a feature selection approach for solving the recognition problem in a multimodal biometric system combining both left and right palmprints of the same subject. In particular, the fusion of two or more traits, at feature-level, results in a long feature vector that needs large storage space, makes the execution time of the recognition task very long, and may include redundant and irrelevant features that can affect the recognition accuracy. To overcome these problems, feature selection is performed using genetic algorithms (GAs) and backtracking search algorithm for a comparison purpose. The experimental results show the usefulness of feature selection, especially the use of genetic algorithms, on the robustness of the multimodal biometric system as regards the feature vector length and run-time reduction, and the significant increase of the recognition rate.

Abundant research has been done on the improvement of the security and trustworthiness of biometric systems. The aim of this paper is to demonstrate the image key-point extraction technique and establish its uniqueness for biometric identification. Ear features comes out to be one of the important biometric systems, which prove to have great potential, in identifying humans in the real world applications. In this work, key-point based matching and recognition is done using SIFT (Scale Invariant Feature Transform) technique. This approach extracts features from images of distinctive invariant. These images are utilized to perform consistent matching between various objects (ear). The key-points are invariant to image scale and hence can provide good matching over a wide range of images. The distinctive features have been matched correctly using the proposed technique and tested on a large database of ear images. This study helps in establishing that the experimental results show improvements in recognition accuracy.

Applications of automated number plate recognition (ANPR) technology in the commercial sector has developed rapidly in recent years. The applications of ANPR system such as vehicle parking, toll enforcement, and traffic management are already widely used but not in Indonesia today. In this paper, the Labeling algorithm and a fully connected neural network are used to create an ANPR system for vehicle parking management in Universitas Multimedia Nusantara, Indonesia. The system is built using Java and the Android SDK for the client and PHP for the server. The proposed ANPR system is targeted for Indonesian civilian number plate. Testing shows that the ANPR system has been implemented successfully. Evaluation of the system gives a precision value of 1 and a recall value of 0.78. These values are obtained with hidden layer nodes of 75, 85, and 95. These number of hidden nodes delivers an F-score of 0.88 with the accuracy of 88%.

In modern manufacturing industry featured with automation and flexibility, the intelligent machine tools management is essential for the workshop. In this work, we proposed a novel machine tools recognition system for classifying 3D models. A common and standard 3D tool database is constructed. The hybrid networks of Convolutional Neural Networks (CNNs) and Extreme Learning Machine (ELM) are developed for multiple view based 3D shape recognition. This framework utilizes the composited advantages of deep CNN architecture with the robust ELM auto-encoder feature representation, as well as the fast ELM classifier. The experimental results shows that it outperforms other methods which are using the manually specified 3D feature descriptors.

High precision image pattern recognition technology based on large data background is an important part of computer application technology. With the rapid development of digital image processing technology and pattern recognition technology, image pattern recognition technology has been applied more comprehensively in various fields. This paper briefly introduces image processing technology and its new development trend, focusing on image recognition technology and new development.

In this paper simple localization method is proposed which uses radio-frequency identification (RFID) system working with low frequencies. The position of RFID tag is obtained by the recognition system installed on a small cartesian robot. In the robot head a RFID reader is installed. It was assumed that the used components for localization should be inexpensive. In the localization the inversion of the typical method based on proximity is applied. The difference is that the position of moving reader is not examined by tags, but position of tags is recognized by moving reader. The method was able to obtain precision in a range of a few millimeters. Two variants of collecting detection points and two variants of calculating position were examined. It was examined how the distance between the reader and the tag influences on position error, and what is the influence of head velocity on the detection accuracy.

Face recognition is an important security task. We propose a high-level method to solve this problem: a permutation coding neural classifier (PCNC). A PCNC with a special feature extractor for face image recognition systems is a relatively new method that has been tested with good results to classify real environment images (such as larvae of various types and hand-made elements). As baseline methods, a support vector machine (SVM) and the iterative closest point (ICP) method are selected for comparison. We applied these methods to gray-level images from the FRAV3D, FEI and LWF (Labeled Faces in the Wild) face databases. We aggregated various distortions for the initial images to improve the PCNC. We analyze and discuss the obtained results. For LWF database we have investigated experimentally three different cases. In the first, the recognition process was based on images of the whole faces. In the second case, the recognition process was based on fragment (eye-eyebrow) images. In the third case, the recognition process was based on fragment (mouth-chin) images. The results are presented. We describe recognition system for the Colorado potato beetles based on RSC neural classifier.

The area of human activity recognition is gaining momentum with the rise of smart appliances towards tracking and monitoring human behavior system. Till last decade, there have been various works being carried out towards building such a robust system that has led its way to commercial products too. However, after an in-depth investigation, it was found there is a far way to go in order to build up a true and dependable recognition system. Therefore, the proposed system introduces a novel framework meant for human activity recognition system with the sole target to enhance the precision factor in the identification process. A simplified feature extraction process has been introduced in this work that after being subjected to ensemble-training approach is found to improve the identification performance significantly. The study outcome shows better accuracy as well as good system performance.