Application Of Image Annotation Based On Doppler Effect

With the explosive growth of digital images and the increasing demand of image retrieval,the automatic tagging service of picture attracts more and more people's attention,identifying one or more specific people in the picture is a common demand in the social field,such as Facebook Google Picasa and so on,they all have tagging service for the picture.Tagging information of friends in the picture can be better to share happiness with friends and it can also quickly retrieve a friend's photo in the massive pictures.The usual labeling methods are manual tagging and automatic tagging of picture.The meaning of manual tagging is the user manually tagging the name and the other information in the picture.The method is so inefficient,heavy workload and prone to mistakes.Traditional tagging service of picture is based on image analysis algorithms,the method has high complexity and unsatisfactory accuracy.The image tagging application based on the Doppler effect proposed in this paper is that people will automatically mark the information of people in the picture after taking the picture,and the accuracy and complexity of the picture tagging are greatly improved.In this paper,the Doppler shift of acoustic wave is used to identify the characteristics of the receiving devices.The frequency of the received sound waves is different because of the location of the different receiving devices is different.According to the size of sound frequency received by the device to determine the relative position of each device and then acquire the relative position in the picture of the person who holding the device,so that we can tag the name of the photographer in the picture or other information.To achieve this function need to address two major challenges: 1)when take photos with some people in the crowd,how to distinguish this part of the people appear in the photo and other people do not appear in the picture.2)How to accurately associate each tag information with the corresponding person in the picture.To address these two challenges,we propose a solution based on the Doppler effect applied to the phone.First,After the photographer finishes taking picture,the photographer moves the mobile phone(the sending end)and sends the sine wave of 20 kHz through the loudspeaker.The photographed people's phone(receiving end)samples the received acoustic signal at a rate of 44.1 kHz and passes through the 10-order Band-pass filter to attenuate frequency components below 19 kHz and above 21 kHz.And then after the filtered signal under-sampling processing,the sampling of the audio samples at 10 ms interval,75% overlap rate of window framing processing,and then calculate each frame of data by the FFT algorithm,select The maximum frequency of all the frames as the sound frequency received by the camera's mobile phone,and at last the photographed by the phone reply frequency and users' name to the camera's mobile phone through the network to achieve mark function automatically according the message the sender receive.Based on the description above,this paper designs and realizes the high precision detection Doppler frequency shift and the picture automatic tagging system,and realizes the scheme on 7 android mobile phones,and also takes pictures in different real scenes.Experimental results show that in the range of 3 meters,the accuracy of more than 85%.by the picture automatic tagging and compare with the existing system Picasa.The comparison result show that the system and Picasa system have good performance in the correct rate and false rate indicators and the system performance is better than the Picasa system in the recall rate.