For all you live streamers/filmers and all those that try and remain anonymous when filming please note the way you shake gives you away.
So there are many ways that can get you recognised from your recording there is the standard activity recognition. This involves looking at what’s being filmed, a glimpse of hands, object recognition, limited to controlled environments and limited to types of activities. This type of profiling is standard and takes man power, then there is the social interaction, where they are looking, who is looking where, recognising “important” people that can be used to create the scene giving more information about the person recording, but now it has all gone and got technical and Shmuel Peleg and Yadid Hoshen at Israel ‘ s Hebrew University of Jerusalem have developed a software that can identify film makers in seconds making people like, base jumpers who often trespass large corporations to complete their mission and protesters that can be of interest to the authorities easier to identify. This can also work against the authorities with more police/ private security wearing body cameras they will no longer be able to cover things up by exchanging cameras with colleagues.
So if it is a fixed camera things like the height of the camera, the gait of the person wearing the camera and your shake can identify you with in seconds. The video is easier to match when it is short term recording and Shmuel Peleg and Yadid Hoshen at Israel collected footage from a sample of 34 people wearing go pro’s on baseball caps and ran it through an algorithm that recognised motion signatures to a particular person. This was achieved by dividing frames into non-overlapping blocks. Movement of these blocks between frames reveals the vertical and lateral motion of the camera. The system identified the camera wearer by movement alone 88% of the time and only 12 seconds of footage where needed to make the identification
This also raises issues with the use and sharing of camera’s they could well have software in the future that recognises the recorder and only allows that person to use the equipment. This may well be integrated with in the software in things like Google glasses and even for insurance purposes if a camera is stolen it could be blocked and no longer usable.
As with all forms of control and surveillance it is a constant battle of technologies against our personal freedoms and already there is talk of ways to beat this software when uploading by using programmes that can change the optical flow pattern without significantly changing the content. It may even be possible to tamper with video to make it look as if it were filmed by somebody else.
What we have to do now is to make sure the police/private security are wearing body cameras that they can not turn on and off at their own discretion and that all body cams are person specific.
Citizen journalism is needed for real unbiased reporting and is an essential part of our freedoms in revealing the oppression of the state. So we need to counter this somehow. Open source of programs to be used whilst uploading our footage is a place to start.
So back to filming …….