The analysis of temporal behavioral patterns of home network users can reveal important information to Internet Service Providers (ISPs) and help them to optimize their networks and offer new services (e.g., remote software upgrades, troubleshooting, energy savings). This study uses time series analysis of continuous traffic data from wireless home networks , to extract traffic patterns recurring within, or across homes, and assess the impact of different device types (fixed or portable) on home traffic. Traditional techniques for time series analysis are not suited in this respect, due to the limited stationary and evolving distribution properties of wireless home traffic data. We propose a novel framework that relies on a correlation-based similarity measure of time series , as well as a notion of strong stationarity to define motifs and dominant devices. Using this framework, we analyze the wireless traffic collected from 196 home gateways over two months. The proposed approach goes beyond existing application-specific analysis techniques, such as analysis of wireless traffic, which mainly rely on data aggregated across hundreds, or thousands of users. Our framework, enables the extraction of recurring patterns from traffic time series of individual homes, leading to a much more fine-grained analysis of the behavior patterns of the users. We also determine the best time aggregation policy w.r.t. to the number and statistical importance of the extracted motifs, as well as the device types dominating these motifs and the overall gateway traffic. Our results show that ISPs can exceed the simple observation of the aggregated gateway traffic and better understand their networks.