作者: Enrique Garcia-Ceja , Ramon Brena
DOI: 10.1016/J.PROTCY.2013.04.031
关键词:
摘要: Abstract In the last years, simple activity recognition through wearable sensors has been achieved successfully, however complex is still challenging. Simple activities may just a few seconds, e.g., walking, running, resting, etc. whereas involve combination of former and they from minutes to several hours. this work long-term performed modeled as distribution represented histogram. For experiments, raw histograms were used for task then we added an additional step which consists extracting features over histogram applying threshold reduce noise. This resulted in increase on classification accuracy.