Abstract: Falling is one of the main causes of casualties in the elderly, every year about 40 million people over the age of 65 fall accidentally. To improve the accuracy in human fall detection, a fall detection algorithm based on acceleration sensor and barometer in a smart phone is proposed, the algorithm is an improved support vector machine (SVM). Firstly, it uses the SVM to train the training set to obtain a weak 2-classifier (including the optimal hyperplane and support vector set), and then calculates the distance from the sample to the optimal hyperplane. If the distance is greater than the given threshold, the tested sample would be classified with SVM. Otherwise, the K-nearest-neighbor classifier (KNN) method will be used. In addition, in the KNN method, the distance between the eigenvectors is calculated using the standard Euclidean distance. Simulation results show that compared with the non-optimized support vector machine algorithm, this algorithm can effectively improve the fall detection accuracy and smartphones can be placed casually.

Key words: fall detection, SVM, KNN, SVM_KNN, Matlab