Deep-Learning-Based Multivariate Time-Series Classification for Indoor/Outdoor Detection

Recently, the topic of indoor outdoor detection (IOD) has seen its popularity increase, as IOD models can be leveraged to augment performance numerous Internet Things and other applications. aims at distinguishing in an efficient manner whether a user resides or environment, by inspecting cellular phone sensor recordings. Legacy attempt determine user’s environment comparing measurements some threshold values. However, we also observe our experiments, such exhibit limited scalability, their accuracy poor. Machine learning (ML)-based aim removing this limitation, utilizing large volume train ML algorithms classify environment. Yet, most existing research, temporal dimension problem is disregarded. In article, propose treating multivariate time-series classification (TSC) problem, explore various deep (DL) models. We demonstrate that TSC approach used monitor predict changes state, with greater compared conventional approaches ignore feature variation over time. Additionally, introduce new DL model for TSC, exploiting concept self-attention atrous spatial pyramid pooling. The proposed framework exploits only low power consumption sensors infer it outperforms state-of-the-art models, yielding higher combined smaller computational cost.