Low-power sensor interfaces for detection of sporadic events

doctoral thesis

• Title (croatian): Senzorska sučelja niske potrošnje za detekciju sporadičnih događaja
• Author: Marko Gazivoda
• Mentor: Vedran Bilas (mentor)
• Committee member: Vedran Bilas (predsjednik povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; (Department of Electronic Systems and Information Processing); Zagreb
• Defense date and country: 2022, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Electrical Engineering
• Universal decimal classification (UDC): 004 - Computer science and technology. Computing. Data processing

Abstract

The focus of this thesis was on wake up sensor interfaces used for lowering sensor node power consumptions. The interfaces’ intended application was detection of sporadic transient events in the lower acoustic spectrum (up to 2.5 kHz), lasting up to a few seconds. The passing speedboat event was chosen as a representative event. Acoustic signal features were analyzed, and a set of criteria was developed to determine the feature’s applicability in low power interfaces for passing speedboat detection. Based on the criteria, the spectro temporal decomposition and level crossing rate interface architectures were explored further. Interface prototypes were developed and functionally tested. The results showed both architectures were applicable for passing speedboat detection, with true positive detection rates exceeding 90%. They also showed the level crossing rate architecture was operational with two times lower inputs, required 40% less area and components to implement and had over three times lower power consumption. As a solution for the spectro temporal decomposition architecture’s envelope detector chokepoint, two feature extractors were proposed and developed, one utilizing an electrically and the other a mechanically switched inductor. The feature extractors achieved 43% and 72% lower power consumptions and had four and two times lower minimal input voltage, respectively, compared to the conventional spectro temporal decomposition interface, demonstrating the advantages of the switched inductor in low power sensor interfaces for event detection.

Abstract (croatian)

Predmet doktorske disertacije bila su senzorska sučelja niske potrošnje za detekciju sporadičnih kratkotrajnih pojava (do par sekundi) u donjem dijelu akustičkog spektra (do 2.5 kHz). Reprezentativna pojava od interesa bila je prolazak glisera. Analizirane su značajke akustičkih signala s aspekta primjenjivosti u sučeljima niske potrošnje za detekciju glisera. Detaljnije su analizirane arhitekture bazirane na vremensko frekvencijskoj dekompoziciji i na stopi prolaska kroz zadanu razinu. Obje su se arhitekture pokazale primjenjivima za detekciju glisera, uz točnost detekcije preko 90%. Arhitektura bazirana na stopi prolaska kroz zadanu razinu se pokazala funkcionalnom s dvostruko nižim ulaznim naponima i trostruko nižom potrošnjom te 40% manjom površinom i brojem komponenata potrebnim za implementaciju. Kao rješenje problema ispravljača kao uskog grla arhitekture bazirane na vremensko frekvencijsko dekompoziciji, predložena su dva nova sklopa za ekstrakciju značajki, bazirana na elektronički i mehanički preklapanoj zavojnici, koji su primjenjivi uz 43%, odnosno 72% posto nižu potrošnju te mogu raditi s četiri, odnosno dva puta nižim ulaznim naponima od konvencionalne arhitekture bazirane na vremensko frekvencijskoj dekompoziciji, ukazujući na prednosti upotrebe preklapane zavojnice u sensozorskim sučeljima niske potrošnje. Doprinosi disertacije: Arhitekture senzorskih sučelja niske potrošnje za odabrane značajke signala niske frekvencije generiranih sporadičnim pojavama Analiza primjenjivosti ispravljača s preklapanom zavojnicom u detekciji slabih signala

• Keywords: always-on wake-up interfaces; interface architectures; acoustic event detection; acoustic signal features; switched inductor feature extractor; passive electromechanical feature extractor
• Keywords (croatian): uvijek aktivna senzorska sučelja za buđenje; arhitekture sučelja; detekcija akustičkih događaja; značajke akustičkog signala; ekstraktor značajki s preklapanom zavojnicom; pasivni elektromehanički ekstraktor značajki
• Language: english
• URN:NBN: urn:nbn:hr:168:292495
• Promotion: 2022
• Study programme: Title: Computer Science; Study programme type: university; Study level: postgraduate; Academic / professional title: Doktor znanosti (Doktor znanosti)
• Type of resource: Text
• Extent: 164 str. : graf. prikazi
• File origin: Born digital
• Access conditions: Open access
• Created on: 2022-05-23 13:26:50