Dual Mode Data Acquisition and Analysis Based on Deep Learning for Smart Home Networks

As the sharp increase in data traffic, except for security, home network and data usage will become two important issues in smart home in the future. On the one hand, aiming to solve the problem of wall jack and room coverage in smart home networks, we develop a set of dual-mode data acquisition device, which supports wired power line communication (PLC) and wireless dual-mode communication. Our PLC module is with the advantage of being pluggable on the debugging board, which can be replaced with a module that supports different PLC protocols as required, such as broadband PLC or narrowband PLC. On the other hand, aiming to solve massive data usage problem acquired by smart home facilities, we propose a data analysis method that integrates the automatic encoder scheme and the
$K$
-means algorithm. On this basis, we propose a Conv-LSTM-Attn algorithm that is a prediction model of convolutional long- and short-term memory network based on attention mechanism. Through the field test, it illustrates that our system can realize accurate long-term data acquisition and recording, dual-mode control from indoor to long-distance away, the autonomous backhaul of the collected data with low-latency and long-distance transmission. Besides, it can highly flexibly configure various sensor modules to achieve smart home management and control. Moreover, through the analysis of actual sampling data, the proposed Conv-LSTM-Attn algorithm outperforms the existing machine learning schemes in smart home in terms of achieving the prediction values.