The invention discloses a wearable human body falling detection device which is composed of embedded type multi-sensor hardware and falling detection software, wherein human body posture change is measured by the embedded type multi-sensor hardware through an acceleration sensor and a tilt angle sensor, and the falling detection software is used for judging whether the abnormal behavior of falling occurs; when falling occurs, a notice is given to a far end for medical assistance through the General Packet Radio Service (GPRS) and the Global Position System (GPS); a measured human body posture signal is filtered by the falling detection software and then multiple characteristic quantities are extracted; training of multiple SVM parameters and the weighting coefficient of each SVM of the diversity AdaBoost-SVM integrated algorithm is carried out on the characteristic quantities to form an SVM integrated classifier; finally data acquired in real time are input into the falling detection software for detection. According to the wearable human body falling detection device, the embedded diversity AdaBoost-SVM classifier can be used for providing help when an emergency occurs to the aged who are outside, and therefore the device has broad market prospect and great application value.