Evaluation of Single HMM as a Pre-Impact Fall Detector Based on Different Input Signals

Accurate near-fall (pre-impact) detection has always been a goal of any fall prevention system. It allows for the early interception before an injury has been sustained by the user. This paper looks at the application of hidden Markov model (HMM) as a near-fall detector with different input signals from multiple IMUs. More input signals and post-processing may increase sensitivity but they also increase computation time and detection latency. This study aims to determine the simplest, fastest and most accurate HMM-based pre-impact fall detection algorithm. 5 IMUs placed at the torso, thighs and shanks are used for data collection. Multiple combinations of acceleration, gyroscope, orientation and quaternion are used as inputs to HMM, without any feature extraction or complex post-processing. Results show that the algorithm is capable of fall detection at about 200 ms after fall initiation with 90% sensitivity and 92% precision. The best performing IMU placements are at the torso and thigh. In addition, the computational latency of this algorithm can be as fast as 0.45 ms.

[1]  Karen L Troy,et al.  Trunk kinematics and fall risk of older adults: translating biomechanical results to the clinic. , 2008, Journal of electromyography and kinesiology : official journal of the International Society of Electrophysiological Kinesiology.

[2]  Jeffrey M. Hausdorff,et al.  Evaluation of Accelerometer-Based Fall Detection Algorithms on Real-World Falls , 2012, PloS one.

[3]  Benjamin Georgi,et al.  The General Hidden Markov Model Library : Analyzing Systems with Unobservable States , 2004 .

[4]  Geoffrey E. Hinton,et al.  SMEM Algorithm for Mixture Models , 1998, Neural Computation.

[5]  Ralf Salomon,et al.  iFall - a new embedded system for the detection of unexpected falls , 2010, 2010 8th IEEE International Conference on Pervasive Computing and Communications Workshops (PERCOM Workshops).

[6]  A K Bourke,et al.  Evaluation of a threshold-based tri-axial accelerometer fall detection algorithm. , 2007, Gait & posture.

[7]  W C Hayes,et al.  Disturbance type and gait speed affect fall direction and impact location. , 2001, Journal of biomechanics.

[8]  Haiyong Luo,et al.  A fall detection algorithm based on pattern recognition and human posture analysis , 2011 .

[9]  Patrick Boissy,et al.  A smart sensor based on rules and its evaluation in daily routines , 2003, Proceedings of the 25th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (IEEE Cat. No.03CH37439).

[10]  Xiaole Cui,et al.  Towards HMM based human motion recognition using MEMS inertial sensors , 2009, 2008 IEEE International Conference on Robotics and Biomimetics.

[11]  L. Rubenstein.  Falls in older people: epidemiology, risk factors and strategies for prevention. , 2006, Age and ageing.

[12]  Xiang Chen,et al.  A Framework for Daily Activity Monitoring and Fall Detection Based on Surface Electromyography and Accelerometer Signals , 2013, IEEE Journal of Biomedical and Health Informatics.

[13]  Gang Zhou,et al.  Accurate, Fast Fall Detection Using Gyroscopes and Accelerometer-Derived Posture Information , 2009, 2009 Sixth International Workshop on Wearable and Implantable Body Sensor Networks.

[14]  Joydeep Ghosh,et al.  Evolving Gaussian Mixture Models with Splitting and Merging Mutation Operators , 2016, Evolutionary Computation.

[15]  Adrian E. Raftery,et al.  How Many Clusters? Which Clustering Method? Answers Via Model-Based Cluster Analysis , 1998, Comput. J..

[16]  D. Winter,et al.  Strategies for recovery from a trip in early and late swing during human walking , 2004, Experimental Brain Research.

[17]  S. Robinovitch,et al.  Video capture of the circumstances of falls in elderly people residing in long-term care: an observational study , 2013, The Lancet.

[18]  Shuwan Xue,et al.  Portable Preimpact Fall Detector With Inertial Sensors , 2008, IEEE Transactions on Neural Systems and Rehabilitation Engineering.

[19]  Yunjian Ge,et al.  HMM-Based Human Fall Detection and Prediction Method Using Tri-Axial Accelerometer , 2013, IEEE Sensors Journal.

[20]  Maarit Kangas,et al.  Comparison of low-complexity fall detection algorithms for body attached accelerometers. , 2008, Gait & posture.

[21]  Shehroz S. Khan,et al.  X-Factor HMMs for Detecting Falls in the Absence of Fall-Specific Training Data , 2014, IWAAL.