Hardware components | ||||||
 |
| × | 1 | |||
In a world where smartphones gain more and more relevance it’s almost mandatory to always carry the device in the pocket. Permanent online availability belongs to one of the notable standards of living because many people cannot imagine a life without it. This is reflected in the constant increasing data traffic whether it’s streaming media like videos on Youtube or music using a provider like Spotify as well as the increasing number of text, voice or video messages sent through various channels. All this shows, that people have developed a serious dependence to their smartphone.
To keep this availability on the socially expected level, an application shall be provided which helps in a frequent but painful situation. Meant is the situation where people put their phones away in the car. An app was developed that focuses on this problem where drivers forget the most important technical device in this place.
Based on sensors commonly provided by modern smartphones and activated by users it is possible to precisely determine everyday's situations. For detecting that the smartphone is placed in its holder simple metrics like average and standard deviation along short time intervals like 1/2 second are enough. The following graphs show four typical situations like driving, walking, engine in neutral and engine off for the amplitude, acceleration and rotation sensors.
Due to these observations thresholds for the classified datasets can be determined (threshold results are given in the code section). After applying the obtained thresholds for driving, engine in neutral and engine off on a test dataset it can be seen that the classification is quite good. However, not each sensor is able to precisely determine any classified situation.
To ensure that the overall situation is classified correctly, such as the user can be reminded to pick up his phone after driving, are the individual sensor classifications combined by a fully state machine (code provided). (4) The state unclassified is reached if the rotation sensor provides unclassified data. This will mostly happen if the user already uses its smartphone which makes the reminder useless. (3) Driving is reached if acceleration and amplitude values indicate the driving situation. (2) The idle engine state is reached if the amplitude sensor data matches and acceleration sensor is classified as engine off or engine idle. The combination of the acceleration classifications is needed because, as already shown in the graphs, are these classes not distinguishable. (1) For the transition to the state nothing is the combination of the acceleration classes the same as well as the amplitude classification must return engine off.
With some more steps like sensor debouncing it is possible to build an app which makes "forgetting the forgotten smartphone after car driving" possible.
Comments