Crowdsourcing Based Image Collection And Training System For Expression Recognition

Face expression recognition refers to the machine through the access to human facial expression information,through a certain algorithm to determine the psychological state of mind.The academic point of view is the facial expression information for feature extraction and classification process.Facial recognition is an important cross-cutting topic across a wide range of disciplines such as image processing disciplines,machine vision disciplines and psychology.But the novel and efficient algorithms are supported by a large number of image data.There is much demand for the number of calibrated expression data samples.The lack of emotional sample image is the biggest problem that restricts the development of expression recognition algorithm.Crowdsouring is a new form of labor production.In recent years,crowdsourcing has spread widely into the Internet.Although it is a relatively new technology,but the package technology has a well-designed incentive mechanism,a large number of all walks of life crowds of workers,a variety of flexible forms of organization of production and efficient information flow.These advantages make the crowdsourcing technology has produced many compelling success stories.In this paper,we use the mobile terminal APP and the social platform to share the acquisition and calibration of the expression sample image.Based on the Andrews platform,we develop an emotional sample image acquisition system,which attracts through the APP and the social platform.Many users join in the collection and calibration of emotional image images.The system determines the final calibration of the expression based on the user's nominal number of votes.The system in the design of the use of social networking platform is to spread quickly,spread the advantages of high efficiency,short time to collect the case of high accuracy of the human face expression data and its calibration information.The system continues to input and train the collected samples,evolutionary expression recognition algorithm,for the special expression is still a great recognition accuracy.Finally,this paper tests the whole system,including the functional test of the core function module,the performance test of the expression recognition rate,and the continuous evolution of the CNN algorithm by using the massive emotion sample images collected by the system to test the system identification effect.The test completes the collection,recognition,calibration and storage of the expression data,and combines the social function with the system function.The system collects a large number of emotional sample image data that will greatly contribute to the study of facial recognition.