Investigation of signal processing techniques for acoustic reflectometry

(1)

INVESTIGATION OF SIGNAL PROCESSING TECHNIQUES FOR ACOUSTIC

REFLECTOMETRY

KAPIL DEV TYAGI

CENTRE FOR APPLIED RESEARCH IN ELECTRONICS

INDIAN INSTITUTE OF TECHNOLOGY DELHI

SEPTEMBER 2016

(2)

©Indian Institute of Technology Delhi (IITD), New Delhi, 2016

(3)

INVESTIGATION OF SIGNAL PROCESSING TECHNIQUES FOR ACOUSTIC

REFLECTOMETRY

by

KAPIL DEV TYAGI

CENTRE FOR APPLIED RESEARCH IN ELECTRONICS

Submitted

in fulfillment of the requirements of the degree of Doctor of Philosophy

to the

INDIAN INSTITUTE OF TECHNOLOGY DELHI

SEPTEMBER 2016

(4)

Dedicated to MY PARENTS Smt. Raj Bala Tyagi Shri Arvind Kumar Tyagi

&

TEACHERS

(5)

Certificate

It is certified that the thesis “Investigation of Signal Processing Techniques for Acoustic Reflectometry”which is being submitted by Mr. Kapil Dev Tyagi for the award of the degree ofDoctor of Philosophyin the Centre for Applied Research in Electronics of the Indian Institute of Technology Delhi is a record of the students own work carried out by him under our joint supervision and guidance. The matter embodied in this thesis has not been submitted to any other University or Institute for the award of any degree or diploma.

Professor Arun Kumar Professor Rajendar Bahl

Centre for Applied Research in Electronics Centre for Applied Research in Electronics Indian Institute of Technology Delhi Indian Institute of Technology Delhi

New Delhi-110 016 (India) New Delhi-110 016 (India)

Date:

New Delhi

i

(6)

Acknowledgements

First and foremost I would like to thank my Ph.D. supervisors, Prof. Arun Kumar and Prof. Rajendar Bahl for giving me this wonderful opportunity. Their invaluable guidance and support throughout the research work, as well as their painstaking effort in proof reading the drafts, are greatly appreciated. I enjoyed the freedom that I was given during the course of this research work, that allowed me to explore and learn more than I otherwise might have. Their willingness to listen to my presentations and their patient approach to rectifying the mistakes and suggesting improvisations have been the main factors that helped shape this thesis.

This thesis could not be completed without the support of my wife Gunjan Tyagi.

Her invaluable companionship, warmth, strong faith in my capabilities and me, has always helped me to be more assertive at work. Her support, optimistic and enlightening discussions, encouragement, love and extraordinary patience have made it possible to complete this research work.

I wish to thank my seniors, Dr. Lokesh S S for his mentoring and for being a supportive friend. I am grateful to my friends and fellow Ph.D. students specially Mr.

B. Suresh and Pushpendra Singh, for their great company. I also thank the non-teaching staff of the Centre for Applied Research in Electronics for their support and assistance.

I express my deep sense of reverence and profound gratitude to my parents and grandparents, family members and specially to my brother Vaibhav for their uncondi- tional love and support. I am very thankful to my parents-in-law Smt. Munesh and Shri D.K. Tyagi for their love, continuous support and blessings. I also express my sincere thanks to my brothers-in-law for his progressive discussions, motivations and healthy

(7)

company. Finally, I would like to thank my daughter Era, son Adamya and Atharv to keep me stress free throughout this research work. I owe every success in my life to them.

Kapil Dev Tyagi

iv

(8)

Abstract

There are numerous applications that can be accomplished in a non-invasive man- ner using acoustic reflectometry such as gas pipeline leakage detection, underwater sedi- ment classification, marine life monitoring, oil exploration, soil analysis, study of layered media like snow, soil etc. One of the applications of acoustic reflectometry in which we are interested is the stratigraphy studies of layered media.

The thesis begins with a discussion on some of the important issues related to the design of a portable and robust acoustic reflectometry system including its hardware and software. A prototype portable battery operated acoustic reflectometry system has been developed, calibrated and tested in the laboratory in a full anechoic room facility, so that it can be used for stratigraphy studies in the field. A novel signal processing algorithm using inverse filtering is developed for artifact compensation and reliable detection of the received signals.

In the next topic, the CTFM technique for the analysis of the received signal in an acoustic reflectometry system at low SNR is explored. It is observed that the range resolution in CTFM is limited by the usable bandwidth of the probe signal. The super resolution Root-MUSIC algorithm for analysis of the output in CTFM processing is studied and proposed to achieve higher range resolution. It has been demonstrated by an experiment conducted in a full anechoic room that the super resolution based analysis gives higher range resolution than the conventional DFT based analysis and is therefore useful for the analysis media which have severe constraint on the usable bandwidth. Next, the DD-CTFM technique is implemented and it is observed that the claim of getting higher range resolution using DD-CTFM does not work in practice. This

(9)

has been explained both mathematically and through simulation study. A novel signal processing algorithm is proposed to overcome the range resolution limitation of the DD- CTFM technique. The proposed enhanced range resolution dual demodulator continuous time frequency modulation (ER-DD-CTFM) algorithm is tested in both simulation and practical experiments. The range resolution obtained using the ER-DD-CTFM algorithm is ten times better than the bandwidth constrained limit of CTFM technique.

Next, a directional loudspeaker based novel approach for the detection of the received signal from a layered media at audio frequencies is proposed. It has been demonstrated using a theoretical example of snow stratigraphy that this approach can make the analysis simpler. The proposed technique is also tested in a laboratory experiment for the detection of the speed of sound in foam medium.

Finally, field experiments are conducted to determine the attenuation constant and the speed of sound in snow medium. A number of field experiments were conducted using the designed acoustic reflectometry system for snow stratigraphy studies. The claims made by previous researchers about snow stratigraphy using acoustic reflectometry were found to be inconsistent. Several issues in using acoustic reflectometry for snow stratigraphy studies that have not been discussed in the research literature are presented.

Finally, the conclusions of the thesis and possible directions for future research are given.

vi

(10)

List of Figures

1.1 Time frequency plot of transmitted and received signal. . . 10

2.1 Acoustic reflectometry system block diagram. . . 21

2.2 Layered structure for acoustic probing. . . 22

2.3 Different types of connector cables used in acoustic reflectometry system. 32 2.4 Portable acoustic reflectometry system block diagram. . . 33

2.5 A proposed tripod frame arrangement to hold the speaker and the microphone. . . 34

2.6 A tripod frame arrangement designed to hold the speaker and the microphone . . . 35

2.7 Metal plate for fixing at the base of tripod legs. . . 35

2.8 Chain system of tripod legs. . . 36

2.9 Bipod metal frame to hold portable acoustic reflectometry system. . . 37

2.10 Backpacks containing the acoustic reflectometry system components. . . 38

2.11 Time synchronization of transmission and reception. . . 39

3.1 Cross-correlation function between transmitted and received LFM signal. 48 3.2 Basic principle of frequency domain method. . . 49

3.3 DFT magnitude plot of low pass filtered signal. . . 49

3.4 Time frequency plot and blind time problem of CTFM technique. . . 51

3.5 Experiment setup to study the direct pickup response of the acoustic reflectometry system. . . 52

3.6 DFT magnitude plot of the low pass filtered signal. . . 53

(17)

3.7 DFT magnitude plot of the low pass filtered signal with loudspeaker from Sony. . . 54 3.8 DFT magnitude plot of the difference frequency signal with loudspeaker

from M-Audio. . . 55 3.9 Inverse filter design for artifact compensation. . . 56 3.10 DFT magnitude plot for the compensated received signal using filter de-

signed in time domain. . . 56 3.11 DFT magnitude plot for the compensated received signal using filter de-

signed in frequency domain. . . 58 3.12 DFT magnitude plot for the compensated received signal that is recorded

separately from the signal used for the inverse filter design. . . 58 3.13 Illustration of different paths from the loudspeaker to the microphone. . . 59 3.14 DFT magnitude plot for the compensated received signal with chirp signal

of final frequency of 2.1 kHz. . . 60 3.15 DFT magnitude plot to study the effect of temperature on inverse filter

performance in artifact suppression. . . 61 3.16 DFT magnitude plot of the low pass filtered signal with inverse filter and

person standing near the setup. . . 62 3.17 Test for 25 degrees tilt of the loudspeaker on inverse filter performance. . 63 3.18 DFT magnitude plot of the low pass filtered signal with loudspeaker tilted

by 25 degrees towards the microphone. . . 63 3.19 DFT magnitude plot of the low pass filtered signal with loudspeaker and

microphone separation is increased by 10 cm. . . 64 3.20 DFT magnitude plot to study the reflection from a sheet. . . 65 3.21 Illustration of multiple reflections between the media and the loudspeaker

box. . . 65 3.22 Interface structure setup 1 to study the reflections. . . 67 3.23 The DFT magnitude plot shows the peaks corresponding to the reflections

from setup 1 of the interface structure. . . 67

xiv

(18)

3.24 Interface structure setup 2 to study the reflections. . . 68 3.25 The DFT magnitude plot shows the peak corresponding reflections from

the setup 2 of interface structure. . . 69 4.1 Experiment setup for simulation study. . . 75 4.2 DFT of signal consists of two sinusoids of 30 Hz and 33.7 Hz of duration

270 ms. . . 76 4.3 Bias and standard deviation in Hz in the error of estimation of difference

frequencies 30 Hz and 33.7 Hz using Root-MUSIC with sampling rate varied from 800 Hz to 2400 Hz and correlation matrix size proportionately varied from 40x40 to 120x120 for SNR=20 dB. . . 77 4.4 Bias and standard deviation in Hz in the error of estimation of difference

frequencies of two echoes using Root-MUSIC with sampling frequency fixed at 1 kHz and correlation matrix size varied from 60x60 to 120x120 (only number of rows of the square matrix is displayed on x-axis). . . . 78 4.5 Bias and standard deviation in Hz in estimation of frequencies of two

signals with varying relative amplitude in dB at SNR=40 dB. . . 79 4.6 Bias and standard deviation in Hz in estimation of frequencies of two

signals with varying difference between two frequencies at SNR=40 dB. . 81 4.10 Bias and standard deviation in Hz in estimation of frequencies of two

signals with varying difference between two frequencies at SNR=20 dB. . 82 4.11 Bias and standard deviation in Hz in estimation of frequencies of two

signals with varying difference between two frequencies at SNR=5 dB. . 82

(19)

4.12 Bias and standard deviation in Hz in estimation of frequencies of two signals with randomly varying phases at SNR=40 dB. . . 83 4.13 Bias and standard deviation in Hz in estimation of frequencies of two

signals with randomly varying phases at SNR=20 dB. . . 83 4.14 Bias and standard deviation in Hz in estimation of frequencies of two

signals with randomly varying phases at SNR=10 dB. . . 84 4.15 Bias and standard deviation in Hz in estimation of frequencies of two

signals with increasing signal duration (initial chirp duration is of 360 ms) and reducing frequencies at SNR=40 dB. . . 85 4.16 Bias and standard deviation in Hz in estimation of frequencies of two

signals with increasing signal duration (initial chirp duration is of 360 ms) and reducing frequencies at SNR=30 dB. . . 85 4.17 Bias and standard deviation in Hz in estimation of frequencies of two

signals with increasing signal duration (initial chirp duration is of 360 ms) and reducing frequencies at SNR=25 dB. . . 86 4.18 Block diagram for lab experiments. . . 87 4.19 Setup for lab experiments. . . 88 4.20 DFT of signal consisting of two sinusoids of 30 Hz and 31.57 Hz of duration

270 ms. . . 88 5.1 Receiver processing for DD-CTFM technique. . . 93 5.2 Time-frequency plot and DD-CTFM output in the ideal case without

phase discontinuity. . . 94 5.3 DFT of the output signal in DD-CTFM processing without phase discon-

tinuity problem . . . 95 5.4 Time-frequency plot and DD-CTFM processing: difference between ideal

and desired waveform (2^nd from bottom) and actual waveform (bottom). 96 5.5 DFT magnitude plot of the output signal in DD-CTFM processing with

phase discontinuity problem. . . 97

xvi

(20)

5.6 Addition of the outputs of the two channels without and with phase cor-

rection. . . 103

5.7 DFT magnitude plot of the output signal in DD-CTFM processing with no phase correction applied. . . 104

5.8 DFT magnitude plot of the output signal in DD-CTFM processing with phase correction applied for difference frequency signal of 32 Hz. . . 105

5.9 Results from proposed algorithm at 20 dB SNR. . . 113

5.10 Results from proposed algorithm at 5 dB SNR. . . 113

5.11 The amplitudes of the difference frequencies corresponding to the echoes with recorded signal, the circle on the frequency axis indicates the ex- pected difference frequencies. . . 115

6.1 Block diagram of acoustic reflectometry system. . . 122

6.2 Single layer structure model. . . 124

6.3 Multiple layer media model. . . 126

6.4 Direct recording from the loudspeaker to the microphone. . . 128

6.5 Recording from the loudspeaker to the microphone with foam layer in the path. . . 129

6.6 Layer properties determination method using directional loudspeaker. . . 130

7.1 Experiment setup for measurement of attenuation constant. . . 136

7.2 Attenuation constant measurement system arrangement in the field. . . . 136

7.3 Stepped FM signal recorded at 5 cm depth from snow surface. . . 137

7.4 Stepped FM signal recorded at 25 cm depth from snow surface. . . 137

7.5 Attenuation constant versus frequency. . . 138

7.6 Envelope of cross-correlation function signal. . . 140

7.7 Speed of sound as a function of frequency. . . 140

7.8 Field trials for snow stratigraphy studies using acoustic reflectometry. . . 141

(21)

7.9 DFT magnitude plot of difference frequency signal of the field recorded signal with artifact compensation (explained in Section 3.9) corresponding to chirp signal 1. . . 142 7.10 DFT magnitude plot of difference frequency signal of the field recorded

signal with artifact compensation corresponding to chirp signal 2. . . 143 7.11 Simulation setup for simulation study. . . 145 7.12 Cross-correlation between transmitted and received MLS. MLS is gener-

ated with shift register length of 16 bits and duration 1 second. . . 146 7.13 Cross-correlation between transmitted and received MLS. MLS is gener-

ated with shift register length of 7 bits and periodically repeated to create a total of 44100 samples for a duration 1 second. . . 147 7.14 Cross-correlation between transmitted and received MLS. MLS is gener-

ated with shift register length of 3 bits and periodically repeated to create a total of 44100 samples for a duration 1 second. . . 147 7.15 Cross-correlation between transmitted and received MLS. MLS is gener-

ated with shift register length of 7 bits and stretching each bit, so the total duration is 1 second. . . 148 7.16 Cross-correlation between transmitted and received MLS. MLS is gener-

ated with shift register length of 3 bits and stretching each bit, so the total duration is 1 second. . . 149 7.17 Cross-correlation between transmitted and received MLS (passed from a

band pass filter of bandwidth 1 kHz) with shift register length of 16 bits and duration 1 second. . . 153

xviii

(22)

List of Tables

2.1 Specifications of components of the acoustic reflectometry system. . . 31 2.2 Different types of connector cables used in acoustic reflectometry system. 32 3.1 Probe waveforms used in acoustic reflectometry. . . 45 3.2 Specifications of the experiments. . . 52 3.3 Distances traveled by the different components of the received signal and

corresponding difference frequencies for setup 1. . . 68 3.4 Distances traveled by the different components of the received signal and

corresponding difference frequencies for setup 2. . . 69 4.1 Distinguishable relative signal amplitude with a maximum bias of ±0.2

Hz in estimation of difference frequencies at various SNR. . . 81 5.1 Parameters for calculation of phase mismatch in DD-CTFM processing. . 99 5.2 Specifications of the simulation. . . 112 5.3 True and estimated difference frequencies in Hz using ER-DD-CTFM at

SNR=20 dB and SNR=5 dB. . . 113 5.4 Specifications of the simulation. . . 114 5.5 Reflected and multiple reflected distances traveled by echoes and corre-

sponding difference frequencies. . . 116 5.6 Comparison in the range resolution improvement by ER-DD-CTFM over

Root-MUSIC analysis based CTFM and DD-CTFM technique. . . 117 7.1 Acoustic signal attenuation constant at different frequencies. . . 139

(23)

List of Abbreviations and Symbols

ADC Analog to Digital Converter.

DAQ Data Acquisition System.

dB Decibel

dBA A-weighting, an environmental noise Measurement.

SNR Signal-to-Noise ratio.

SPL Sound Pressure Level.

SWE Snow Water Equivalent.

LFM Linear Frequency Modulation.

FMCW Frequency Modulated Continuous Wave.

CTFM Continuous Time Frequency Modulation.

DD-CTFM Double Demodulator Continuous Time Frequency Modulation.

ER-DD-CTFM Enhanced Resolution Double Demodulator Continuous Time Frequency Modulation.

MUSIC Multiple Signal Classification.

ESPRIT Estimation of Signal Parameters Via Rotational Invariance.

f₁ Starting frequency of a chirp signal.

f₂ Final frequency of a chirp signal.

µ Rate of frequency change of a chirp signal.

ρ Layer density.

ϕ Continuity constant.

Γ Reflection coefficient.

hr Impulse response of snowpack.

xxi

(24)

h_a Attenuation response of snowpack.

φ_k Porosity of k^th snow layer.

α_k Tortuosity ofk^th snow layer.

ρ_ice Ice density.

ζrar Attenuation (in dB) in reflected signal due to reflection coefficient.

ζrat Attenuation (in dB) in transmitted signal due to reflection coefficient.

ζ_sa Attenuation (in dB) due to spreading loss.

F Far field distance for an acoustic signal.

λ Wavelength of the signal.

Λ(_T^t) Symmetric triangle function.

G(k) Transfer function of inverse filter.

Investigation of signal processing techniques for acoustic reflectometry

INVESTIGATION OF SIGNAL PROCESSING TECHNIQUES FOR ACOUSTIC

REFLECTOMETRY

KAPIL DEV TYAGI

CENTRE FOR APPLIED RESEARCH IN ELECTRONICS

INDIAN INSTITUTE OF TECHNOLOGY DELHI

SEPTEMBER 2016

INVESTIGATION OF SIGNAL PROCESSING TECHNIQUES FOR ACOUSTIC

REFLECTOMETRY

by

KAPIL DEV TYAGI

INDIAN INSTITUTE OF TECHNOLOGY DELHI

SEPTEMBER 2016

Certificate

Acknowledgements

Abstract

Contents

List of Figures

List of Tables

List of Abbreviations and Symbols