ISOEN 2011|Jin Huang|CSE@TAMU Overview Introduction What is active - PowerPoint PPT Presentation

ISOEN 2011|Jin Huang|CSE@TAMU

Overview • Introduction – What is active sensing? – Infrared Fabry ‐ Perot interferometry • Methods – Non ‐ negative least squares – Multi ‐ modal search – Wavelength selection • Results – Chemical dataset – System behavior – Comparison with passive sensing • Discussion ICASSP 2013 | � lab | CSE@TAMU 2

What is active sensing? • Perception is an active process – “We not only see but we look, we not only touch we feel” ‐ J.J. Gibson • Active sensor vs. active sensing – Active sensor : a device that transmits energy in order to make measurements • E.g., radar, sonar – Active sensing : a control strategy that dynamically adapts the sensor’s configuration as it interacts with the environments • E.g., changing camera viewpoints ICASSP 2013 | � lab | CSE@TAMU 6

What is active sensing? • Analogy – Guessing games • 20 questions, Pictionary, Battleship, Yes and no, Hangman Question Answer Vague Is your character real? Yes Isyour character male? Yes Is he alive? Yes Is he an actor? No Is he linked with sports? No Is he a musician? Yes Is he more than 50 years old? Yes Does he play the guitar? Yes Is he American? Yes Does he wear headgear? Yes Does he have messy hair? Yes Specific Is it Bob Dylan? Yes ICASSP 2013 | � lab | CSE@TAMU 9

What is active sensing? • Analogy – Guessing games • 20 questions, Pictionary, Battleship, Yes and no, Hangman – Two seemingly conflicting problems at the same time • Find the right answer • Ask the right questions – Advantage • Incorporate decision early in the signal processing pipeline • Reduce sensing costs ($, energy, time, computing power) • Lower sensor requirements ICASSP 2013 | � lab | CSE@TAMU 10

Prior work Chemical Identity Concentration Sensor Reference Single Unknown Fixed MOX Gosangi et al. 2010 IEEE Sensors J Single Unknown Unknown FPI Huang et al. 2012 (continuous) IEEE Sensors J Mixture Known Unknown MOX Gosangi et al. 2013 (discrete) S&A B: Chemical Mixture Unknown Unknown FPI Huang et al. 2013 (continuous) ICASSP ICASSP 2013 | � lab | CSE@TAMU 11

Infrared absorption spectroscopy ����� Gas cell IR spectrometer IR source � � � � � � � � log �� � � � � � ICASSP 2013 | � lab | CSE@TAMU 12

Fabry ‐ Perot interferometer � � � � � ����� Gas cell FPI IR source � � � � � �� � , … , � �� � � � ��� � d ICASSP 2013 | � lab | CSE@TAMU 13

IR absorption • Properties – Absorption spectrum is linear (Beer’s law) � � � � � � � � ��� Absorption spectrum � Absorption spectrum of Concentration of each each c hemical � � chemical � � of the mixture � �� � ⋮ � � � � � �� + � �� ⋮ � � � � � �� ICASSP 2013 | � lab | CSE@TAMU 18

Traditional mixture analysis • Problem definition – Estimate � , given � (all wavelengths) � �� � � … � � … � � �. �. : � � 0 � ( � � � ) � ( � � 1 ) � ( � � 1 ) Wavelengths � � Chemicals ICASSP 2013 | � lab | CSE@TAMU 22

Active mixture analysis • Problem definition – Estimate � , but measuring one wavelength at a time � �� � � … � � … � � �. �. : � � 0 � ( � � � ) � ( � � 1 ) � ( � � 1 ) Wavelengths � � Chemicals ICASSP 2013 | � lab | CSE@TAMU 23

Active mixture analysis • Problem definition – Select the best wavelengths in � to solve �� � � � �� � � … � � … � � �. �. : � � 0 � ( � � � ) � ( � � 1 ) � ( � � 1 ) Wavelengths � � Chemicals ICASSP 2013 | � lab | CSE@TAMU 24

Active mixture analysis • Interpretation – Each wavelength equals one element in � � �� � � … � � … � � �. �. : � � 0 � ( � � � ) � ( � � 1 ) � ( � � 1 ) Wavelengths � � Chemicals ICASSP 2013 | � lab | CSE@TAMU 25

Active mixture analysis • Interpretation – Which allows us to use an additional row in � � �� � � … � � … � � �. �. : � � 0 � ( � � � ) � ( � � 1 ) � ( � � 1 ) Wavelengths � � Chemicals ICASSP 2013 | � lab | CSE@TAMU 26

Active mixture analysis • Interpretation – Each new wavelength adds a new row in � � �� � � … � � … � � �. �. : � � 0 � ( � � � ) � ( � � 1 ) � ( � � 1 ) Wavelengths � � Chemicals ICASSP 2013 | � lab | CSE@TAMU 27

Active mixture analysis • Issue – The underlying linear system may be under ‐ determined � � � � � � � … � � … � � �. �. : � � 0 � � ( 2 � 1 ) � � ( 2 � � ) � ( � � 1 ) Wavelengths � � Chemicals ICASSP 2013 | � lab | CSE@TAMU 28

Active mixture analysis • Solution – Assume that � is sparse � � � � � � � … � � … � � �. �. : � � 0 � � ( 2 � 1 ) � � ( 2 � � ) � ( � � 1 ) Wavelengths � � Chemicals ICASSP 2013 | � lab | CSE@TAMU 29

Active mixture analysis • Subset selection – Select 1 ‐ 2 elements in � � � � � � � � … � � … � � �. �. : � � 0 � � ( 2 � 1 ) � � ( 2 � � ) � ( � � 1 ) Wavelengths � � Chemicals ICASSP 2013 | � lab | CSE@TAMU 30

Active mixture analysis • Interpretation – One element in � implies one column in � � � � � � � � … � � … � � �. �. : � � 0 � � ( 2 � 1 ) � � ( 2 � � ) � ( � � 1 ) Wavelengths � � Chemicals ICASSP 2013 | � lab | CSE@TAMU 31

Active mixture analysis • The selection is not unique – A combinatorial problem ⇒ search � � � � � � � … � � … � � �. �. : � � 0 � ( � � 1 ) Chemicals ICASSP 2013 | � lab | CSE@TAMU 33

Unimodal candidate selection ICASSP 2013 | � lab | CSE@TAMU 37

Multimodal candidate selection � � � Iterative deepening Iterative deepening memory ‐ bounded memory ‐ bounded heuristic search heuristic search ICASSP 2013 | � lab | CSE@TAMU 43

Wavelength selection Unknown spectra b � Maximum variance Feature Sensor Selection Select � ��� Projection MM ‐ NNLS � � � �� � � � solver Solutions: � � , � � �, … � �� ICASSP 2013 | � lab | CSE@TAMU 44

Case study • Dataset – 100 chemicals from NIST WebBook (randomly chosen) – Wavelength range: 3 � 11.5�� – Downsampled to 660 spectral lines – Added 2% Gaussian noise • Setup – 3 chemicals mixture (sparsity 3%) – Search space 100 � � 10 � – We consider up to 10 � alternate paths ICASSP 2013 | � lab | CSE@TAMU 45

• Step 1 -3 x 10 Ground truth Measurement Projection Selected feature 6 Absorption t=1 4 2 0 3 4 5 6 7 8 9 10 11 Wavelength (  m) l 2 error 10 20 30 40 50 60 70 80 90 100 Candidate error ranking ICASSP 2013 | � lab | CSE@TAMU 46

• Step 2 -3 x 10 Ground truth Measurement Projection Selected feature 6 Absorption t=2 4 2 0 3 4 5 6 7 8 9 10 11 Wavelength (  m) l 2 error 200 400 600 800 1000 1200 1400 1600 1800 Candidate error ranking ICASSP 2013 | � lab | CSE@TAMU 47

• Step 3 -3 x 10 Ground truth Measurement Projection Selected feature 6 Absorption t=3 4 2 0 3 4 5 6 7 8 9 10 11 Wavelength (  m) l 2 error 500 1000 1500 2000 2500 3000 3500 4000 Candidate error ranking ICASSP 2013 | � lab | CSE@TAMU 48

• Step 4 -3 x 10 Ground truth Measurement Projection Selected feature 6 Absorption t=4 4 2 0 3 4 5 6 7 8 9 10 11 Wavelength (  m) l 2 error 500 1000 1500 2000 2500 3000 3500 4000 Candidate error ranking ICASSP 2013 | � lab | CSE@TAMU 49

• Step 5 -3 x 10 Ground truth Measurement Projection Selected feature 6 Absorption t=5 4 2 0 3 4 5 6 7 8 9 10 11 Wavelength (  m) l 2 error 1000 2000 3000 4000 5000 6000 Candidate error ranking ICASSP 2013 | � lab | CSE@TAMU 50

• Step 6 -3 x 10 Ground truth Measurement Projection Selected feature 6 Absorption t=6 4 2 0 3 4 5 6 7 8 9 10 11 Wavelength (  m) l 2 error 1000 2000 3000 4000 5000 6000 7000 800 Candidate error ranking ICASSP 2013 | � lab | CSE@TAMU 51

• Step 7 -3 x 10 Ground truth Measurement Projection Selected feature 6 Absorption t=7 4 2 0 3 4 5 6 7 8 9 10 11 Wavelength (  m) l 2 error 1000 2000 3000 4000 5000 6000 Candidate error ranking ICASSP 2013 | � lab | CSE@TAMU 52

• Step 8 -3 x 10 Ground truth Measurement Projection Selected feature 6 Absorption t=8 4 2 0 3 4 5 6 7 8 9 10 11 Wavelength (  m) l 2 error 1000 2000 3000 4000 5000 6000 7000 Candidate error ranking ICASSP 2013 | � lab | CSE@TAMU 53

• Step 9 -3 x 10 Ground truth Measurement Projection Selected feature 6 Absorption t=9 4 2 0 3 4 5 6 7 8 9 10 11 Wavelength (  m) l 2 error 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 Candidate error ranking ICASSP 2013 | � lab | CSE@TAMU 54