Automatic switching between forward- and rear-facing cameras on smartphones and tablet computers to match the user's intention by analyzing the input to both cameras
Publication Date: 2014-Apr-30
The IP.com Prior Art Database
Disclosed is an automatic switching between forward- and rear-facing cameras on smartphones and tablet computers to match the user's intention by analyzing the input to both cameras.
Page 01 of 2
Automatic switching between forward - and rear-facing cameras on smartphones and tablet computers to match the user 's intention by analyzing the input to both cameras
Smartphones and tablet computers from various manufacturers are now starting to come with a forward-facing camera in addition to the more conventional rear-facing camera. The rear-facing camera is typically more powerful and is meant to be used as the main camera while the forward-facing camera is typically less powerful but is essential for uses such as taking self-portraits and video calling.
Currently, users have to switch between the two cameras manually, typically by tapping on a software control on the capacitive touchscreen. This manual aspect what I wish to improve on.
My idea is that when the user activates the camera functionality, the system will then analyze the input to both cameras to determine which camera the user actually intends to use and then choose that camera automatically. If the user is attempting to take photos of people, objects or landscapes in the conventional way, then the device will detect this and select the rear-facing camera automatically. If the user is attempting to take a self-portrait, looking into the device as if looking into a mirror, then the forward-facing camera will be chosen automatically.
The detection and camera switching can happen dynamically. For example, within the same video call, in the middle of normal chatting using the forward-facing camera, the user can point the smartphone's rear-facing camera towards an object he wants the other party to see and then the rear-facing camera will be switched-to automatically. And when the user resumes normal chatting, the system will then switch back to the forward-facing camera automatically. This transition can happen multiple times within the same video call. The automatic detection and switching provides a seamless user experience.
There are mature fields of theory (such as artificial intelligence, computer vision, machine learning, etc.) which describe techniques (such as neural networks, Bayesian inference, decision tree learning, support vector machines, etc.) that can be used by the system to solve the problem of determining which one of the two cameras is more relevant at any given time. Theory and practice can also guide us on how to tune these techniques for accuracy, like minimizing false positives and negatives, and performance. Here are some key concepts to provide a little context:
- One approach used by some techniques (also known as classifiers) is that a sufficiently complex function of weights and values f(w1v1, w2v2, w3v3, w4v4, ...) can approximate any function (for example, recognizing printed characters, where the input may be a bitmap and the output is recognized text) with the right set of weights.
- Training the system means to use examples, known input and output pairs, to iteratively adjust the weights so that the output error is minimized.