In the last decades the world of embedded systems has become pervasive in our every day lives, and it is still in the heart of its growth; meanwhile the improvement of Automatic Speech Recognition (ASR), Speech To Text (STT) and interactive interfaces based on voice, that are nowadays common on each OS, is changing our conception of human-machine interaction. Many of those technologies, though, are not able to adapt perfectly to acoustically noisy environments yet. The beamforming techniques allow the virtual handling of physical acoustic spaces, permitting the enhancement, or even the isolation, of the acoustic source of interest coming from a certain direction respect to the acoustic interferences coming from other directions. Differential Microphone Arrays, in respect to the other beamforming techniques, seem to be ideal for embedded systems, thanks to properties such as frequency independence, low computational cost, high directivity and integration in small spaces. In this thesis, we introduce some implementations of Differential Microphone Arrays, focusing on first and second order with linear and circular geometries. Through these implementations the goal of isolating the source of interest, preserving its integrity in terms of signal-to noise ratio and frequency response, has proved to be not a trivial challenge, as long as working on embedded systems means avoiding "hard computing"˙ The theory behind Differential Microphone Arrays often omits some of the constraints, limitations and errors introduced by a real implementation system. In this thesis we aim to study and evaluate when these problems interfere significantly and how to overcome and avoid them. In conclusion, the goal of this thesis is to find and evaluate the optimal design of a differential microphone array for embedded systems which has to be robust to noise, high directive and with a frequency response suitable for frequency-broadband signals.
Negli ultimi decenni il mondo dei sistemi integrati è diventato pervasivo nella nostra vita quotidiana, ed è ancora nel cuore della sua crescita; contemporaneamente l’affacciarci ai servizi di Authomatic Speech Recognition (ASR), Speech To Text (STT) e interfacce interattive basate sulla comunicazione verbale orale, ormai presente in tutti i principali sistemi operativi, sta cambiando la nostra concezione di interazione uomo-macchina. Tuttavia molte di queste tecnologie non sono ancora perfezionate ad adattarsi ad ambienti acusticamente rumorosi. Le tecniche di Beamforming, consentono la gestione virtuale dello spazio fisico acustico, permettendo di poter enfatizzare o isolare una sorgente acustica proveniente da una certa direzione e di poter annullare le interferenze acustiche provenienti da altre direzioni. Gli array differenziali di microfoni si distinguono tra le tecniche di beamforming per la loro indipendenza dalla frequenza, semplicità computazionale, alta direttività e per l’integrazione in piccoli spazi, caratteristiche che rendono questo tipo di tecniche ideali per i sistemi integrati. Nel corso di questa tesi, vengono presentate diverse implementazioni di array differenziali di microfoni, con attenzione su primo e secondo ordine con geometrie lineari e circolari. Attraverso queste implementazioni, isolare una sorgente acustica di interesse, preservandone la sua integrità in termini di rapporto segnale-rumore e di risposta in frequenza, si è rivelata una sfida non banale, soprattutto perché, lavorando su sistemi integrati, verranno evitate tecniche troppo "pesanti" da un punto di vista computazionale. Nella teoria sugli array differenziali di microfoni, molti dei vincoli, limitazioni ed errori introdotti da un’implementazione su un sistema reale, vengono trascurati o omessi, e in questa tesi si vuole studiare quando questi problemi sono significativi per la resa del beamformer, e valutare come evitarli e risolverli. In definitiva lo scopo di questa tesi è quello di trovare e valutare il design ottimale di un array differenziale di microfoni per sistemi integrati, che sia robusto al rumore, altamente direttivo e con una risposta in frequenza ideale per segnali "broadband".
An embedded circular differential microphone array
FLORIDIA, GIUSEPPE
2014/2015
Abstract
In the last decades the world of embedded systems has become pervasive in our every day lives, and it is still in the heart of its growth; meanwhile the improvement of Automatic Speech Recognition (ASR), Speech To Text (STT) and interactive interfaces based on voice, that are nowadays common on each OS, is changing our conception of human-machine interaction. Many of those technologies, though, are not able to adapt perfectly to acoustically noisy environments yet. The beamforming techniques allow the virtual handling of physical acoustic spaces, permitting the enhancement, or even the isolation, of the acoustic source of interest coming from a certain direction respect to the acoustic interferences coming from other directions. Differential Microphone Arrays, in respect to the other beamforming techniques, seem to be ideal for embedded systems, thanks to properties such as frequency independence, low computational cost, high directivity and integration in small spaces. In this thesis, we introduce some implementations of Differential Microphone Arrays, focusing on first and second order with linear and circular geometries. Through these implementations the goal of isolating the source of interest, preserving its integrity in terms of signal-to noise ratio and frequency response, has proved to be not a trivial challenge, as long as working on embedded systems means avoiding "hard computing"˙ The theory behind Differential Microphone Arrays often omits some of the constraints, limitations and errors introduced by a real implementation system. In this thesis we aim to study and evaluate when these problems interfere significantly and how to overcome and avoid them. In conclusion, the goal of this thesis is to find and evaluate the optimal design of a differential microphone array for embedded systems which has to be robust to noise, high directive and with a frequency response suitable for frequency-broadband signals.| File | Dimensione | Formato | |
|---|---|---|---|
|
2015_12_Floridia.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: "An Embedded Circular Differential Microphone Array"
Dimensione
8.03 MB
Formato
Adobe PDF
|
8.03 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/116321