| （英） |
Micro Electro Mechanical Systems (MEMS) sensors are promising for IoT applications because they are small and available at low prices. In general, the data observed by a MEMS sensor is temporarily stored in the register and needs to be read out by the host computer before it is updated. Therefore, a delay in reading register may result in the failure of data extraction. In contrast, reading the register twice before update results in the acquisition of redundant data. Therefore, it is desired that the sampling interval by the host computer exactly equals to the update period of the register. For that reason, estimation methods of this update period are herein discussed which is based on the observed data. To achieve this purpose, consecutive pairs of adjacent observed data are compared and binary values are assigned depending on whether they are equal or not. The obtained data sequence is regarded as a collection of binary random variables. Using this model, the update period is estimated by means of maximum likelihood estimation. Examples of applying this method and another to actual MEMS sensor data are presented. |