A hand gesture recognition system with CNN开题报告

1. 研究目的与意义（文献综述包含参考文献）

At present, humanmachine interaction is very important for operating the machines in a remote manner by the commands which are received from humans. In this regard, gestures are playing an important role in operating the machine at a distant mode. The machines capture the gestures from the human and recognize it for operating the machines. communication or transfer of data in between human and human is really easy and understandable. But when it comes to human and machine its really difficult because even machine knows all the languages humans can speak and understand, they cannot communicate with that knowledge or data To perceive motions, distinctive highlights, for example, handmade spatiotemporal descriptors and enunciated models were utilized. As signal classifiers, concealed Markov models, contingent irregular fields and bolster vector machines (SVM) have been broadly utilized. Notwithstanding, vigorous order of signals under broadly fluctuating lighting conditions, and from various subjects is as yet a testing issue.Computer-human communicationrefers to the way how the human communicate to the computer/machine, and since the machine is not useful until a human trains the machine for a particular task. There are mainly 2 characteristics that will be checked when developing a man-machine communication model as mentioned in: machines performance and usage. The Model performance refers to how well the machines are performing to communicate with the human and usage refers to weather all the provided functionalities are performing according to the development.Hand gestures can be static or dynamic . Static hand gestures are otherwise known as hand postures and are formed of various shapes and orientations of hands without representing any motion information. Dynamic hand gestures are constituted by a sequence of hand postures with associated motion information . Besides the static and dynamic gestures, the gestures of human are also classified into online and offline gestures. The offline gestures operate the icons on the machine, and they are not able to alter the position of the items in the menu or system. The online gestures operate the icons in the machine to different positions or inclinations . The online gestures are very much useful in real time machine operating systems than the offline gestures. Bayes classifier and support vector machine (SVM) methodologies for gesture recognition. These methods did not support large training dataset, and it also required high number of training samples. This drawback is eliminated by proposing CNN classifier .proposing CNN classifier It does not require high number of samples in training mode, and the complexity level of this algorithm is low. The novelty of this proposed work is to implement deep learning algorithm in hand gesture recognition system with novel segmentation technique.The gestures are different types of modes as static and dynamic.1- The static gestures do not change their position, while the machine is operated,2- The dynamic gestures change their positions during the machine is operatedHence, the identification or recognition of dynamic gestures is very important than the static gestures .Hand postures mainly constitute the fingerspelling of the sign language vocabulary, which are used for the letter by letter signing of names, place names, age, numbers, date, year and words that doesnt have predefined signs in the vocabulary . Visual interfacing using hand postures have also received wide acceptance in varied application fields (human computer interaction (HCI) , human robot interaction (HRI) , virtual reality systemsand medical procedures ) as it avoids the physical contact with the traditional interfacing devices. Thus automatic hand posture recognition has been a hot research area and many works exist on the same using vision based approaches and electronic signal based approaches . Among those, the vision based approaches seem to be more user friendly and convenient than others when considering the complexity of data acquisition process.a- Initially, the camera, which is connected with machine, captures the gestures which are generated by humans. b- Second The background of the detected gestures is removed, and the foreground of the gesture is captured. c- Third The noises in the foreground gesture are detected and removed by filtering techniques.d- finally These noise removed gestures are compared with pre-stored and trained gestures for verifying the sign of the gestures

2. 研究的基本内容、问题解决措施及方案

1. Research task of this projectThe goal is the human hand gestures are detected and recognized using CNN classification approaching our paper, the human hand gestures are detected and recognized using CNN classification approach. The system includes four parts:a- acquiring a gesture sample, b- gesture sample processing,c- run-time gesture recognition d- and a control system. For the first step in acquiring a sample from user, use Open CVo activate the system camera to get the original image . The plotting and how the system can see the image and the model is made to perform operation accordingly. And the downloaded data set will be having different data sets. The system is made to work on each and every one of the sample data that is provide by the kaggle dataset repository. The path flow of the processing is provided 1. Research methods of this project The model uses a convolutional neural networks for hand gesture recognition. in this project we willclassified that gesture recognition system into mainly getting the input image/data from the camera or from any input providing devices. This section introduces the way for operating the data provided from the user from cameras or other input devices. The layers will be created and trained with many sample data and will try with a real time data for the checking it with many random inputs. The dataset used for implementation of this model is a CNNusable dataset of images. It consists of more to 20,000 images with more than 1.2 G.B. This data will be implemented in our model for creating a trained model which is used for recognizing the gestures and for the implementation of those gestures for some operations in systems. This data consists of sample images of Palm, Fist, Fist Moved, Thumb, Index, The data which user will be giving as an input will be processed by a neural network and will be training that data in the CNN model. Once the data is successfully trained it can take sample data from the user and make it recognize and make the model remember and work on the model to perform an operation when the same gesture is invoked.(((((The date we will use old date from some old papers studied the same topic)))))