Indian New Currency Denomination Identification System with Audio Feedback Using Convolution Neural Networks for Visually Challenged and Elderly People

Identification of the denomination of the currency note to pay physically without UPI is the first step of paying to the seller by the consumer. In this project, we have proposed an approach to detect denominations of Indian currency using Convolutional Neural Networks. Computer Vision and object detection is an area of great interest for research in today’s world. It has several applications like detection of defects in machinery, intruder detection, computer vision for code and character recognition among many others. Through the work we have done, we explored something that could be of great help to people in day-to-day life. In this project we have tried to investigate the approaches to detect currency denominations using Convolutional Neural Networks. The objective is to build a model that would be able to detect Indian currency denominations efficiently. Typically the model will be useful for people with vision impairment. The experimental results show that the use of Convolutional Neural Networks is a good way and the model can further be improved if it is trained in such a way that it could also identify the regions of interest.

The currency has a significant meaning in everyday life. Researchers have suggested diverse approaches to improve currency recognition for visually impaired people by implementing portable devices and various image-processing techniques. However, only some of the nearly developed currency recognition systems cater to meet the Sri Lankan currency identification requirement. This paper presents a model that works on a mobile application based on a Convolutional Neural Network using a pre-trained model to identify currency notes in real time from a smartphone camera. Crucial factors in constructing this model are speed, accuracy, and size. A self-collected dataset of 14,000 Sri Lankan currency note images in numerous conditions was used for primary model training. Data augmentation, transfer learning through VGG16, VGG19, ResNet50, DenseNet, MobileNet, NASNet, and EfficientNet pre-trained models using bottleneck features and fine-tuning on top of bottleneck features are the main methods used in this model. This model reached above 98% in terms of testing accuracy using DenseNet121 with fine-tuning by testing 315 partially occulted, rotated Sri Lankan currency notes in different daylight conditions and cluttered backgrounds. The proposed model can recognize the currency note using TensorFlow Lite when it is pointed at the camera. The present research is limited to daylight image capturing to ensure the better performance of the model. The findings of this research would help visually impaired people to recognize Sri Lankan currency notes precisely and with less effort through the simple installation of a smartphone.

Visually impaired individuals face significant challenges in distinguishing between different denominations of Indian currency notes, which hinders their ability to perform routine tasks such as shopping and financial management. Traditional methods of currency recognition, such as tactile features or optical scanners, are often burdensome and inconvenient. This groundbreaking system utilizes a deep learning model developed from a vast dataset of Indian rupee notes captured in diverse environments, including various orientations, stages of wear-and-tear, and lighting conditions, ensuring robust training. By leveraging Convolutional Neural Networks (CNNs), the system ensures a high degree of accuracy in identifying notes and generates distinct audio cues corresponding to each denomination, simplifying the differentiation process for visually impaired individuals. The use of neural networks enables the system to continuously refine its recognition capabilities, adapting and improving its identification methods with additional data, including updated note variations or unexpected scenarios. Upon widespread availability, this technology has the potential to revolutionize how blind individuals in India manage their finances, providing them with increased autonomy and confidence in everyday transactions. The proposed system offers several advantages over traditional currency recognition methods: it is non-invasive, requiring no physical contact with the currency notes, thus enhancing convenience; its portability enables implementation on mobile devices, facilitating access for users on the go; and it prioritizes user-friendliness by providing real-time audio feedback, further simplifying the currency identification process for visually impaired individuals. Overall, this innovative solution represents a significant stride towards empowering the visually impaired community in India, offering newfound freedom and self-assurance in managing their monetary affairs.

Purpose: The purpose of the study was to find out the preferred, accurate, and effective way to identify the new Indian currency note by a blind individual. Methods: A cross-sectional study was conducted in two phases among 30 subjects in a school for the blind. In phase 1, comparison was done in terms of accuracy and the time taken to identify the Indian currency between the old and new currency notes by using the tactile method by the blind individuals. In phase 2, comparison was done same as phase 1, with the addition of two more note identification methods, one was by using Mani note identifier application via smart phone and the other was Notex. The accuracy and time were analyzed and compared for all three used methods for both phases. Results: Thirty blind individuals were included in the study. The mean age of the subjects was 29.43 years (±10.88 years) and the age range was 18–54 years. The total time taken to identify all the currency notes in each category was recorded in seconds. The average time taken with the help of tactile sense was 75.34 s, but the accuracy here was very poor, even many currencies went unidentified. The average time taken with the help of Notex was 136.54 s. The average time taken with the help of Mani mobile note identifier software was 774.07 s. Conclusion: Notex was the most effective method of identifying Indian currency as compared to the other two methods that were used in our study.

Aims: This study aimed to explore the effect of training set diversity on the performance of deep learning models for predicting early gastric cancer (EGC) under endoscopy. Methods: Images of EGC and non-cancerous lesions under narrow-band imaging (ME-NBI) and magnifying blue laser imaging (ME-BLI) were retrospectively collected. Training set 1 was composed of 150 non-cancerous and 309 EGC ME-NBI images, training set 2 was composed of 1505 non-cancerous and 309 EGC ME-BLI images, and training set 3 was the combination of training set 1 and 2. Test set 1 was composed of 376 non-cancerous and 1052 EGC ME-NBI images, test set 2 consisted of 529 non-cancerous and 71 EGC ME-BLI images, and test set 3 was the combination of test set 1 and test set 2. Three deep learning models, convolutional neural network (CNN) 1, CNN 2 and CNN 3 (CNN 1, CNN 2 and CNN 3 were independently trained using training set 1, training set 2 and training set 3, respectively), were constructed, and their performances on each test set were respectively evaluated. One hundred and thirty-eight ME-NBI videos and 17 ME-BLI videos were further collected to evaluate and compare the performance of each model in real time. Results: On the whole, the performance of CNN 3 was the best. The accuracy (Acc), sensitivity (Sn), specificity (Sp) and area under the curve (AUC) of test set 1 in CNN 3 were 87.89% (1255/1428), 90.96% (342/376), 86.79% (913/1052) and 94.60%, respectively. The Acc, Sn, Sp and AUC of test set 2 in CNN 3 were 95% (570/600), 97.92% (518/529), 73.24% (52/71) and 90.93% respectively. The Acc, Sn, Sp and AUC of test set 3 in CNN 3 were 89.99% (1825/2028), 95.03% (860/905), 85.93% (965/1123) and 94.89%, respectively. The performance of CNN 3 was also the best in videos test set. The Acc, Sn and Sp of videos test set in CNN 3 were 91.03% (142/156), 90.58% (125/138) and 94.44% (17/18), respectively. Conclusions: The deep learning model with the most diverse training data has the best diagnostic effect.

This research paper presents the development and evaluation of an Automatic Number Plate Recognition (ANPR) system using Convolutional Neural Networks (CNN) with Regex correction. The aim is to enhance the accuracy and effectiveness of car verification and security processes at First Technical University, Ibadan. The ANPR system was implemented both without Regex correction and with Regex correction. The evaluation results demonstrate significant improvements in the system's performance when CNN with Regex correction is employed. The CNN-based ANPR system achieves a precision of 1.00, recall of 0.90, and F1-score of 0.95 in accurately identifying number plates. These scores indicate increased accuracy and reduce false positives compared to the system without Regex correction. The integration of CNN and Regex correction effectively handles variations and errors in the number plate data, leading to a reliable and efficient car verification process. Future work can focus on further refining the CNN model and optimizing the Regex correction algorithms to enhance the system's accuracy and robustness. The developed ANPR system, utilizing CNN with Regex correction, shows great potential for enhancing car verification and security in various domains, including law enforcement, parking management, and traffic monitoring

Bank currency is our nation's most valuable asset, and in order to cause financial inconsistencies, counterfeit notes that seem like the real thing are introduced into the financial market. During demonetization time it is seen that so much of currency is floating in market. In general, by a human being, it is very difficult to identify forged note from the genuine not instead of various parameters designed for identification as many features of forged note are similar to original one. To discriminate between fake bank currency and original note is a challenging task. So, there must be an automated system that will be available in banks or in ATM machines. To design such an automated system there is need to design an efficient algorithm which is able to predict weather the banknote is genuine or forged bank currency as fake notes are designed with high precision. This paper proposes a CNN algorithm-based fake currency detection model for authenticating Indian currency notes with denominations of 10, 20, 50, 200, and 500. The results are also fairly good and also proposed model has the accuracy of 99.3%.

Currency recognition, or digitization, refers to the process of converting physical currency, such as banknotes and coins, into digital formats. This transformation enhances convenience, security, accessibility, and cost efficiency in financial transactions, which are essential in the modern economy. This paper introduces a Convolutional Neural Network (CNN) model designed for the recognition of Bhutanese paper currency. The model was trained on a dataset comprising various currency types and denominations, achieving a training accuracy of 91 percent and a testing accuracy of 80.5 percent. The architecture consists of three convolutional layers, followed by a dense layer for classification. The findings suggest that CNNs are effective for currency recognition, with the potential for improved accuracy through the expansion of the training dataset.

Currency is of great importance in everyday life and is the lifeline of every transaction. Even with the vast utilization of cards and other electronic payments, paper currency is still extensively used because of its convenience. There are around 10 different types of notes in Indian currency, and every banknote is different from other in dimensions. For instance, colour and pattern, size of the note, and so on. Most of the people practice the usage of Master Cards and Electronic payment and where the money is accepted for ordinary exchanges because of its accessibility. Visually impaired people undergo the difficulty of differentiating the currency papers. Currency Recognition System (CRS) may be effective for the blind and visually impaired people during the process of currency trade. In the proposed method a Currency Recognition System based on Accelerated- KAZE(AKAZE) algorithm is suggested, Accelerated-KAZE (AKAZE) algorithm is based around the concept of nonlinear diffusion filtering. It aims to provide an accurate and efficient alternative to Oriented FAST and rotated BRIEF(ORB). The proposed system is applied to Indian paper currencies. Originally, some of the preprocessing procedures are conducted on the input paper currency image. Then, colour mapping of the input image is done to distinguish colour components from depth. After this, the ROI (Region of Interest) is the essence of the input image. The AKAZE Algorithm is used to find key points and descriptors of the input image. Finally. the detection of currency is done by using brute force matcher technique, it matches descriptors & keypoints of the input image with every descriptor stored in dataset using Hamming Distance Calculation. In a real-world situation, the proposed system playa a vital role in identifying currency images with greater accuracy of 90 % which is achieved and determined. This proposed system aims to make the life of visually impaired people easier.

Accurate traffic trip prediction is an important guarantee for achieving long-term development of intelligent transportation. To cope with the complex and variable factors of elderly people's public transport travel decision-making behavior as well as to improve elderly people's travel, this paper proposes a deep learning-based convolutional neural network for predicting urban elderly people's public transport travel. Taking the plateau city of Lhasa as an example, the data from October 2018-December 2019 of elderly people's bus passes are selected for deep convolutional neural network and regression model modelling respectively, and the results obtained are compared through the model for prediction experiments on January 2020 data. The experimental results show that the mean absolute error (MAE) as well as the root mean square error (RMSE) are reduced by 0.112 and 0.156 compared with the traditional regression prediction model, further indicating that the model has higher prediction accuracy, and is reliable and practical in predicting elderly people's public transport trips.

Following recent advancements in deep learning and artificial intelligence, spoken language identification applications are playing an increasingly significant role in our day-to-day lives, especially in the domain of multi-lingual speech recognition. In this article, we propose a spoken language identification system that depends on the sequence of feature vectors. The proposed system uses a hybrid Convolutional Recurrent Neural Network (CRNN), which combines a Convolutional Neural Network (CNN) with a Recurrent Neural Network (RNN) network, for spoken language identification on seven languages, including Arabic, chosen from subsets of the Mozilla Common Voice (MCV) corpus. The proposed system exploits the advantages of both CNN and RNN architectures to construct the CRNN architecture. At the feature extraction stage, it compares the Gammatone Cepstral Coefficient (GTCC) feature and Mel Frequency Cepstral Coefficient (MFCC) feature, as well as a combination of both. Finally, the speech signals were represented as frames and used as the input for the CRNN architecture. After conducting experiments, the results of the proposed system indicate higher performance with combined GTCC and MFCC features compared to GTCC or MFCC features used individually. The average accuracy of the proposed system was 92.81% in the best experiment for spoken language identification. Furthermore, the system can learn language-specific patterns in various filter size representations of speech files.

As urbanization accelerates, the prevalence of fire incidents leads to significant hazards. Enhancing the accuracy of remote fire detection systems while reducing computation complexity and power consumption in edge hardware are crucial. Therefore, this paper investigates an innovative lightweight Convolutional Spiking Neural Network (CSNN) method for fire detection based on acoustics. In this model, Poisson encoder and convolution encoder strategies are considered and compared. Additionally, the study investigates the impact of observation time steps, surrogate gradient functions, and the threshold and decay rate of membrane potential on network performance. A comparison is made between the classification metrics of the traditional Convolutional Neural Network (CNN) approaches and the proposed lightweight CSNN method. To assess the generalization performance of the proposed lightweight method, publicly available datasets are merged with our experimental data for training, which results in a high accuracy of 99.02%, a precision of 99.37%, a recall of 98.75%, and an F1 score of 99.06% on the test datasets.

Among the many deep learning methods, the convolutional neural network (CNN) model has an excellent performance in image recognition. Research on identifying and classifying image datasets using CNN is ongoing. Animal species recognition and classification with CNN is expected to be helpful for various applications. However, sophisticated feature recognition is essential to classify quasi-species with similar features, such as the quasi-species of parrots that have a high color similarity. The purpose of this study is to develop a vision-based mobile application to classify endangered parrot species using an advanced CNN model based on transfer learning (some parrots have quite similar colors and shapes). We acquired the images in two ways: collecting them directly from the Seoul Grand Park Zoo and crawling them using the Google search. Subsequently, we have built advanced CNN models with transfer learning and trained them using the data. Next, we converted one of the fully trained models into a file for execution on mobile devices and created the Android package files. The accuracy was measured for each of the eight CNN models. The overall accuracy for the camera of the mobile device was 94.125%. For certain species, the accuracy of recognition was 100%, with the required time of only 455 ms. Our approach helps to recognize the species in real time using the camera of the mobile device. Applications will be helpful for the prevention of smuggling of endangered species in the customs clearance area.

In the indoor environment, the activity of the pedes-trian can reflect some semantic information, for example, if a user’s activity is recognized as taking elevator, the location of the user is inferred to be in the elevator. These activities can be used as the landmarks for indoor localization. The development of sensor technology enhances the smartphone’s sensing and computational capabilities. Using the smartphones, the activity of the pedestrian can be recognized. Current methods rely on extracting complex hand-crafted features, thus leading to the incapability of real time pedestrian activities identification. In this paper, we propose a real time pedestrian activities recognition method based on deep convolutional neural network. A new deep convolutional neural (CNN) network has been designed to learn the proper features automatically. Experiments show that the proposed method achieves approximately 95.21% accuracy in about 2 seconds in identifying nine types of activities, including still, walk, upstairs, up elevator, up escalator, down elevator, down escalator, downstairs and turning. Besides, we transplant the activity recognition algorithm to smartphones using tensorflow. Moreover, we have built a pedestrian activity database, which contains more than 6 GB data of accelerometer, magnetometer, gyroscope and barometer collected with various types of smartphones. We will make it public to contribute the academic research.

This study investigates the use of convolutional neural networks to automatically detect malaria-infected cells in blood smear images, offering an alternative to manual diagnosis, which depends on specialized professionals and adequate infrastructure. Manual diagnosis is time-consuming and prone to human errors, especially in malaria-endemic regions with limited resources. Automated approaches based on convolutional neural networks provide a promising solution to optimize the diagnostic process and improve access to rapid treatment in remote areas. The research evaluates the performance of different convolutional neural network architectures for malaria diagnosis, following the Cross Industry Standard Process for Data Mining methodology to structure preprocessing, modeling, and model evaluation. Preprocessing involved normalization and data augmentation techniques to enhance sample quality and diversity. Two architectures were compared: a customized convolutional neural network designed to balance computational efficiency and accuracy, and an adapted VGG16, recognized for its advanced image feature extraction capabilities. Both models were trained and evaluated using a robust dataset of blood cell images. The customized network achieved 95.6% accuracy, outperforming similar models in the literature and demonstrating its practicality for low-resource settings. It also exhibited high precision, recall, and F1-score, ensuring reliable and balanced classification of infected and healthy cells. This approach reduces reliance on specialists and advanced equipment, making malaria diagnosis more accessible and efficient. The findings position the customized convolutional neural network as a viable solution for automated malaria diagnosis, combining simplicity with high performance and offering significant potential for improving healthcare in endemic regions.