Description:Field of Invention
The present invention isrelating to a system and method based on the most recent studies, the World Health Organization estimates that there are around nearly 1 million visually impaired persons globally, with 150 million experiencing vision impairment and 39 million being completely blind.Due to the similarities in paper texture and size between the various categories, one of the most challenging problems for people with visual disabilities is the inability to differentiate paper currencies.
The Objectives of this Invention
The invention's major goal is to help the visually challenged, who rely on the feel of the note, are baffled by the new currency. The visually blind discern notes based on their length and width, a skill they've honed through time. The visually impaired will not recognize the new denomination's emblem until they are shown them.Only a few well-educated blind persons are aware of this. The new denomination bills of Rs50, Rs100, Rs200, and Rs500 are causing them problems because they are difficult to distinguish.
For the blind, adjusting to the new currency note sizes is a considerable problem. It takes a lot of experience to recognize notes by their duration, breadth, and tactile marks. They struggle to recognize the new denominations because they are perplexed. Money has been widely used for general exchange due to its usefulness, aside from the immediate usage of Master cards and any other sort of electronic payment methods. In order to solve this, we are introducing Currency identification systems that are solely based on visual analysis.
Background of the Invention
In (US2014/8996355B2), The several applications that are discussed here make multilingual communications easier. Multilingual conversations are made possible by the structures and techniques of various incarnations through various communication channels, such as online chat, email, based on language cellular relationships, admissions to forum posts, posting to social networking services, and similar methods. Text translation between various languages is implemented in certain instances by use of telecommunication systems and procedures. Participants of systems and techniques can be encouraged to report wrong or mistaken interpretations by offering rewards in exchange for these reports. There are detailed systems and procedures for judging the precision of interpretations. In addition, (US2008/9865248B2) The disclosure of methods for better text-to-speech conversion. An digital document's text can be converted into audio output which comprises speech that is related to the text in addition to audio contextual information thanks to better text-to-speech conversion. When speech (spoken text) about a document is output, one aspect gives the listener aural contextual signals. Before converting a document to voice, a study of it may serve as the basis for the audio contextual clues. Another element can produce an audio summary of a file. The user can then be supplied with the audio synopsis for an article so they can hear a summary before the written material to generate spoken text via a conversion from text to speech. In (CN2013/105324811B), There are presented configurations for text conversion from sounds received from a situation. For instance, a speech-converting program accepts audio input via a head-mounted display device's microphone collection in one disclosed implementation. One or more potential faces are discovered from visual data collected from the environment. The target faces the user focuses on are identified using eye-tracking data. To pinpoint the location of an audio input connected to a target face, beam formation technologies are utilized on at least an element of the sound input. Those target audio sources are transformed into text and shown on the head-mounted show device's transparency monitor.
In recent years, (Suriya Singh, et al [2017], IJARIIE,Vol-3 Issue-2,pp-3265-3269), presented visual object recognition on a mobile phone as well as a computer vision-based application for recognizing money bills on a low-end smartphone. For recognition, the given solution employs a visual Bag of Words (BoW) technique. Bill categorization is done with the help of an algorithm that uses iterative graph cuts. The recognition problem is therefore framed as a retrieval challenge for instances.This system is an example of mobile device-based instant recognition. With the help of MATLAB, (Srushti Samantet al. [2020] International Journal Of Advance Scientific Research And Engineering Trends,Volume 5, Issue 3 , pp-31-35),constructed the algorithm for money recognition.The classification of various features is done using hamming distance in this method. The confusion matrix is used to show the number of correctly detected notes as well as any errors that happened throughout the recognition process. This system's output is integrated into the graphical user interface (GUI). This system could be used where there are transactions and currency recognition is required. (Ankush Singh and colleagues et al. [2019] International Journal of Engineering Research & Technology (IJERT),Vol. 8 Issue 12,pp-441-443), created a system that provides a method for verifying Indian rupee notes. The verification of money notes is done using image processing methods. The system explains how to extract various characteristics from Indian rupee notes.The features of the note are extracted using the MATLAB software. The serial number, security thread, identification mark, and Mahatma Gandhi portrait are the metrics examined for detecting a genuine note, and recall and precision are derived using data set findings. (VedasamhithaAbburu et al. [2017] Proceedings of 2017 Tenth International Conference on Contemporary Computing ( IC3), suggested an image processing-based method for recognizing cash notes. The proposed method can be used to identify a banknote's country of origin, as well as its denomination or value.Only paper currencies have been taken into account. This method works by first identifying the country using certain regions of interest, and then determining the denomination value using various characteristics on the note such as form, color, size, typography, and other unique aspects. (Viranchi N. Patel, et al. [2017] ICRISET2017,vol. 2, pp. 67–72), introduced the Canny Edge Detector method for segmentation and classification, which uses the NN pattern recognition tool with 95.6 percent accuracy.The currency recognition methodology is implemented using the Neural Network Pattern Recognition software by extracting the center face value using the canny edge detector method and preparing the database. Finally, using the NN pattern recognition programme, the database is trained and recognition is possible. (Shubham Mittal and Shiva Mittal et. al. [2018] 2018 3rd International Conference On Internet of Things: Smart Innovation and Usages (IoT-SIU),pp. 1-6), proposed a deep learning-based approach for recognizing Indian Currency Rupee note denominations from color photos. A classification system has been developed based on the notion of transfer learning, in which a massive convolutional neural network is used to classify photos from new classes after being pre-trained on thousands of natural images.Pre-processing and augmentation of photographs recorded in various settings and lighting situations are used to create an image dataset. Experiments show that it can be used in the creation of dedicated portable devices for recognising banknote denominations. For classifying banknotes, (Anilkumar B, et al. [2018] International journal of Mechanical Engineering & technology,vol.9,pp-884-891), suggested the k-nearest neighbor's method (k-NN). In the feature space, the information consists of the k-nearest training examples. Whether k-NN is used for classification or relapse determines the yield. The outcome is either true or false, depending on the testing inputs.Text-to-speech is used to inform the client about the estimated value of the paper note, providing a simple interface for the client's convenience. To extract ROI and make the template matching process easier, simple image processing methods including thresholding, noise reduction, edge detection, and segmentation are utilized.
Summary of the Invention
The proposed invention will be helpful for the blind people.Currency identification systems that are solely based on visual analysis, on the other hand, are insufficient. Our solution is based on image processing and automates and secures the procedure.
There are now various programs that can detect the currency; however, they are not up to date with the most recent currency adjustments. They'll run on a high-end phone and have a lot of processing power as well as a user interface. These apps require the use of the internet, which necessitates the use of a third party.
Detailed Description of the Invention
In the realm of money recognition research, there are two trends: scanner-based and camera-based. Scanner-based solutions presume that the entire paper is captured (like scanner). Such methods are appropriate for money counting machinery. Camera-based methods assume that the paper is captured by a camera that may only record a portion of the paper. The scanner-based kind is the subject of the majority of related publications [2-5].It's designed to allow people with visual impairments to take a picture of any portion of the paper with their phone and have the system recognize it and tell them the dollar value. The command of blind patients will be converted through speech to text conversion in this research. Users will be able to provide commands to the machine via speech recognition. The value of the note will also be spoken out using text-to-speech features in this Android application. This application will detect currencies using the Azure custom vision API and a machine learning classification algorithm based on photos captured on the phone.This is a TensorFlow Lite sample application for Android. It uses image classification to classify what it sees in real time from the device's back camera. The TensorFlow Lite Java API would be used to accomplish inference, and the demonstration app classifies frames in real time, displaying the top most likely classifications.
The command of blind patients will be converted through speech to text conversion in this research. Users will be able to provide commands to the machine via speech recognition. The value of the note will also be spoken out using text-to-speech features in this Android application. This application will detect currencies using the Azure custom vision API and a machine learning classification algorithm based on photos captured on the phone.
Built on JetBrains' IntelliJ IDEA software and designed exclusively for Android development, Android Studio is the official integrateddevelopment environment (IDE) for Google's Android operating system.In 2020, it will be accessible as a subscription-based service or as a download for Windows, macOS, and Linux-based operating systems. It takes the position of the Eclipse Android Development Tools (E-ADT) as the primary IDE for developing native Android apps.
In this module, we manually collect data using a camera and various angles of different denominations of cash notes.Based on denominations, the denominations are divided into classes.The training is performed on the gathered picture dataset using the Teachable Machine platform supplied by GOOGLE to obtain a trained model in this module.We download the model in ZIP file after it has been trained on the dataset.In this module, we design the android's look and feel.The software is created in such a way that it is simple to use for visually impaired people.The files for the trained model are placed in the android project folder.Additionally, audio files are supplied for audio creation as an MP3 format output.
Teachable Machine trains and runs the models you create in your web browser using TensorFlow. js, a Javascript toolkit for machine learning. This is an Android sample application for TensorFlow Lite. It employs image classification to classify everything it sees from the device's back camera in real time. The TensorFlow Lite Java API is used for inference. In real time, the sample software classifies frames and displays the top most likely categories. It lets the user pick between a floating point or quantized model, the number of threads to employ, and whether to run on the CPU, GPU, or via NNAPI.
Currency detection for the blind using machine learning can be achieved through a combination of computer vision techniques and deep learning algorithms. The steps involved are:
Dataset collection: Gather a diverse dataset of images of different currencies from various angles, lighting conditions, and backgrounds. Each image should be labeled with the corresponding currency type.Preprocessing: Preprocess the collected dataset by resizing the images, converting them to grayscale, and normalizing the pixel values. This step helps in reducing computational requirements and improving the training process.Feature extraction: Extract meaningful features from the preprocessed images to represent the currency notes. Convolutional Neural Networks (CNNs) are commonly used for this purpose, as they can automatically learn hierarchical representations from the images.Model training: Train a deep learning model, such as a CNN, using the preprocessed images and their corresponding labels. Split the dataset into training and validation sets to evaluate the model's performance during training. The model should learn to identify the unique features of each currency type.Model evaluation: Evaluate the trained model on a separate test dataset to assess its accuracy and performance. Measure metrics such as precision, recall, and F1-score to determine the effectiveness of the model.Deployment: Integrate the trained model into a system or application specifically designed for the blind. The system can be implemented on a portable device, such as a smartphone or a dedicated device, with a camera that can capture images of the currency notes.Real-time currency detection: When a blind person needs to identify a currency note, they can use the deployed system to capture an image of the note. The system will process the image using the trained model and provide audio feedback indicating the currency type.It's worth noting that the success of currency detection for the blind depends on the availability and quality of the dataset, the effectiveness of the feature extraction and deep learning model, as well as the accuracy of the image capture process. Continuous testing, refining, and improvement are crucial to enhance the system's performance and make it more reliable for users.
The TEXT TO SPEECH DETECTION FOR VISUALLY IMPAIRED USING ANDROID STUDIO FOR DETECTING CURRENCY application can have several features to cater to the needs of visually impaired users and assist them in detecting currency. Below are some key features that can be incorporated into the app:

Currency Detection: The app should be able to use the device's camera to capture images of currency notes and then detect and recognize the denomination of the currency from the images.Text-to-Speech (TTS) Integration: Implement the Android Text-to-Speech (TTS) API to convert text into spoken words. This feature is essential for audibly communicating the detected currency denomination to the visually impaired user.User-Friendly Interface: Design an accessible and user-friendly interface with large fonts, high contrast colors, and easily distinguishable buttons/icons. The interface should be optimized for blind or visually impaired users using screen readers.Offline Support: If possible, enable the app to work offline, so users don't always require an internet connection to use the currency detection and Text-to-Speech features.
4Claims &1 Figure
Brief description of Drawing
In the figure which are illustrate exemplary embodiments of the invention.
Figure 1, Architecture of the proposed invention , Claims:The scope of the invention is defined by the following claims:

Claim:
1. A system/method to the identification of various currency notes using CNN algorithms, said system/method comprising the steps of:
a) The system starts up, and the manual collection of data or data is collected by the camera (1). The system (2) will be trained the model using Google.
b) The developed system for deployment (3), needs mobile application development, then we can launch our trained model (4) in the mobile application.
c) The CNN method is used for identifying the currency notes, after that the output will be delivered by audio sounds.
2. As mentioned in claim 1, the camera gathers data feed from the scanning the currency notes and image segmentation, that will be feed into our mobile application.
3. As per the claim 1, the training is performed on the gathered picture dataset using the Teachable Machine platform supplied by GOOGLE to obtain a trained model in this module.
4. As mentioned in claim 1, the model in ZIP file is downloaded after it has been trained on the dataset.Additionally, audio files are supplied for audio creation as an MP3 format output after predicting the currency notes.