Implementation of Embedded Multiple Signal Classification Algorithm for Mesh IoT Networks
| Autor: | Troccoli, Tiago, Pirskanen, Juho, Ometov, Aleksandr, Nurmi, Jari, Kaseva, Ville |
|---|---|
| Přispěvatelé: | Nurmi, Jari, Lohan, Elena-Simona, Torres Sospedra, Joaquin, Kuusniemi, Heidi, Ometov, Aleksandr, Tampere University, Electrical Engineering |
| Rok vydání: | 2022 |
| Předmět: | |
| Zdroj: | International Conference on Localization and GNSS 2022 International Conference on Localization and GNSS (ICL-GNSS) |
| Popis: | Angle-of-Arrival (AoA) methods are an Internet of Things (IoT) application, which could be used, for example, in indoor localization. Anchor nodes have an array of antennas and could send the data via Ethernet cable to the cloud that calculates AoA. However, having cable connections means high installation costs, and constantly transferring big chunks of data over some IoT networks, such as mesh, is energy inefficient. The solution of this paper consists in executing AoA locally in anchor nodes. Thus, the paper presents an implementation of a Multiple Signal Classification (MUSIC) algorithm tailor- made for embedded system devices. It calculates a complex eigendecomposition via an equivalent real formulation. It has a detailed memory analysis of the implemented solution that shows its memory requirements satisfy commercial embedded systems for IoT, such as Nordic semiconductor System-on-Chip (SoC) of nRF52 Series and all their SoCs with direction-finding capability. Experiments show that reducing the floating-point precision to shrink its memory footprint does not impact the accuracy. It also shows that minimizing the execution time of its time-consuming peak-finding operation has a few effects on accuracy. acceptedVersion |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|