Ultra Low-Power Always-On Wake-Up by Pulse Pattern Adaptive Recognition for Long Term Biodiversity Monitoring

• Main Authors: Marzetti, Sebastián, Gies, Valentin, Barchasz, Valentin, Best, Paul, Paris, Sébastien, Barthelemy, Hervé, Glotin, Hervé
• Other Authors: Institut des Matériaux, de Microélectronique et des Nanosciences de Provence (IM2NP), Aix Marseille Université (AMU)-Université de Toulon (UTLN)-Centre National de la Recherche Scientifique (CNRS), Laboratoire d'Informatique et des Systèmes (LIS) (Marseille, Toulon) (LIS)
• Format: Conference Object
• Language: English
• Published: HAL CCSD 2021
• Subjects: Ultra Low-Power; Always-on Wake-up; Pattern Detection; Embedded Artificial Intelligence; Biosonar; pulse train; [SPI.NANO]Engineering Sciences [physics]/Micro and nanotechnologies/Microelectronics; [SDE.BE]Environmental Sciences/Biodiversity and Ecology; [SPI.AUTO]Engineering Sciences [physics]/Automatic; [SPI.TRON]Engineering Sciences [physics]/Electronics; Sperm whale
• Online Access: https://hal.science/hal-03610009; https://hal.science/hal-03610009/document; https://hal.science/hal-03610009/file/2020%20-%20CONF%20IEEE%20-%20Sperm%20Whales%20Detector.pdf

International audience This paper presents a mixed analog-digital alwayson ultra low-power wake-up based on pulse pattern analysis. It is used for triggering a high performance multi-channel recorder only when necessary. Its architecture makes the most of ultralow power analog primitives coupled with an embedded digital low power system for fine tuning the pulse detector in order to maximise its efficiency. Such system allows long term biodiversity study, as most of the bioacoustic energy is pulsed. As an example, a case study demonstrates on real sperm whale biosonar the efficiency of our system. Architecture and features extraction using analog primitives are first detailed, followed by embedded digital implementation of the automatic gain control for the pulse detector. Always-on current consumption of this intelligent wakeup is 14µA, with an area under the ROC curve equal to 75%. This allows an autonomy of 2 years on a single CR2032 battery cell.