Nowadays, the amount of elder people living alone is increasing, with all the risks that it involves, maybe the most dangerous threat that they face is fall off with nobody around to help them. A fall at an advanced age usually leads to consequences like bones and hip fracture, which in addition to the low mobility that these people present, make to stand up impossible for them. This situation can get worse if after the fall a person loses consciousness, making it impossible to contact to a third party for help by means of a mobile phone or something similar. Different solutions have been developed in order to accomplish this problem, but some of them are not realistic enough, for example some video solutions invade our privacy, and those which are based on mobile phones expect the user to go everywhere with it. In this paper, it is proposed a low cost solution based on an 9-axis IMU (Inertial Measure Unit), which counts with an accelerometer, gyroscope and magnetometer that will give us the needed information to build a fall detector supported by machine learning. This system will include a gateway, which will be responsible of the data collection and the most complex computations.