All submissions of the EM system will be redirected to Online Manuscript Submission System. Authors are requested to submit articles directly to Online Manuscript Submission System of respective journal.

Elephant Vocalization Direction of Arrival Estimation and External Factors Affecting Wave Propagation Using Acoustic Sensor Network

M.Mayilvaganan 1, M.Devaki 2
  1. Associate Professor, Department of Computer Science, P.S.G college of Arts and Science, Coimbatore, India
  2. Research Scholar, Department of Computer Science, P.S.G College of Arts and Science, Coimbatore, India
Related article at Pubmed, Scholar Google

Visit for more related articles at International Journal of Innovative Research in Computer and Communication Engineering

Abstract

In our real time experimental analysis the acoustic sensor estimates the specific target (elephant) coverage localization at 3 different types of forests. The proposed method consists of estimating Direction of Arrival Estimation using partial hyperbolic circular array and also identified few external factors affecting outdoor propagation and acoustic source localization in forest. We already identified and proposed the two atmospheric conditions affecting sound propagation (wind speed and wind Direction) variation in source and sensor (near by and far away). Now, we started our work to identify some of the additional external factors affecting source localization in densed forest. Our proposed results revealed the various atmospheric conditions (temperature, humidity and Wind velocity, reverberation (echo) inside the densed forest) affects the accuracy of source localization.

 

Keywords

acoustic sensor; hyperbolic circular array; acoustic soure, localizations

INTRODUCTION

Sound Source Localization (SSL) using Sensor Network application at forest is one of the emerging research topics since the early 1990’s. Acoustic Circular Array is fixed for while covering the specific target in outdoor environment. In our real time implementation the outdoor environment (forests) can be extremely affected by external factors[1]-[3]. Elephant sounds are recorded in sensor board at the same time other sounds emitted by other source are also recorded. We have fixed 8 sensor for our application to estimate direction at forest and 4 Sensors circularly covered around (maximum) 3 kms for estimating the Direction of Arrival and remaining sensors used for recording the Temperature, Wind Speed, Wind direction and Humidity. Sensor positions are arranged as (Top,Bottom,Left,Right or East/West/North/South). Database Management System was processed for Sound recorded at different locations. The file name stored in sensor’s are related to their original direction(East,west,north and south)[4][6]. Due to limited battery power efficiency sensor’s are enable mode if they sense the signal otherwise all the nodes are sleep mode. As soon as sensor detects the signal, acoustic fingerprint(Database)is designed and signal pre processing (A/D conversion, Filtering and Compression) are done. Using the pre processed signal SPIN routing protocol is designed. The purpose of using SPIN protocol is to eliminate redundancy.
In our real time experimental analysis we have faced the following problem and rectified [7]. If two sensor’s store a similar signal in exactly the same time. SPIN routing protocol is designed to eliminate the signal redundancy occurred in sensor. In order to use the battery energy efficient the sensor network operation used here is synchronous protocol[8]-[10]. After estimating the DOA arrival we are identified and work some of the few external factors affecting outdoor propagation in forest. The external factors are (i) temperature changing the speed of sound profile (iii) wind speed and direction changing the speed of sound profile iii)reverberation inside the forest(echo,scattering,relection) iv)diffraction by barriers and other considerations.
The rest of the paper is structured as follows. Chapter 2 discusses the existing approaches for finding the Direction of Arrival. Chapter 3 explains the proposed methodology for DOA estimation. Chapter 4 reveals the experimental results for DOA estimation and Error Analysis for external factor affecting outdoor propagation in forest. Chapter 5 outlines the Results and Discussion. Finally, Chapter 6 concludes the paper.

REVIEW OF EXISTING APPROACHES

Specific target DOA estimation in forest is one of the essential sensor network applications in forest. Estimating DOA for an acoustic source of narrow band sources where origin of the source is far away from the sensor array optimal algorithm are processed in existing techniques such as multiple signal classification (MUSIC) the minimum variance method of Capon , estimation of signal parameters via rotational invariance technique (ESPRIT) and more. Better performance was achieved at higher the SNR[10]-14]. The standard method for estimating DOA is ML method.
Observed parameters are essential for formulating likelihood function. Estimation of ML with respect to all unknown parameters, which may include the source DOA angles, the signal covariance, and the noise parameters by maximizing the likelihood function.There are different optimization techniques available in literature for optimization of ML function like AP-AML, simulated annealing (SA) , genetic algorithms (GA) fast EM and SAGE algorithms and a local search technique e.g. Quasi-Newton methods[13]. The evolutionary algorithms like genetic algorithm , particle swarm optimization and simulated annealing [can be designed to optimize the ML function. Genetic algorithm and particle swarm optimization had already used as a global optimization technique to estimate the DOA for uniform array[15].

PROPOSED METHODOLOGY AND DATA FOR RESEARCH

Problem Identification
Let us consider an circular omni-directional array of n sensor which sense the random signal generated at least one target coverage etc, from the specific target around 3 km’s with the unknown location or Direction of Arrival (DOA), w be the direction of sensor located in the forest, L be the life time of a sensor, dc be the set of directions that cover a target[6][18].
Signal Generation in Hyperbolic Circular array
The prototype for all Hyperbolic Partial Differential equations of one-way wave equations
ut +aux= 0 (1)
where a is a constant, t represents time, and x represents the spatial variable.
Before estimating the Direction of Arrival, the recorded signals are pre processed. After pre processing the digital signal is stored in the Acoustic Fingerprint. Routing Protocols are designed for data transmission between acoustic sensor network. Due to limited battery power efficiency the sensor network are divided in to three modules[13].
(i) Sensor identification and Management
(ii) Routing Protocol
(iii) Network Topology
The following figure represents the signal pre processing results before direction of arrival estimation.

EXPERIMENTAL RESULTS AND DISCUSSIONS

Due to several external factors our DOA accuracy is affected in densed forest. Due to heavy wind speed , the elephant emits the sound far away from the sensor the results are affected by various metrological condition such as(wind speed and temperature) etc. The following figure shows the doa estimation for an stand alone elephant in the palamalai hill. Due to heavy wind speed DOA results are highly affected by error rate.
Factors Affecting Outdoor Sound Propagation In Forest
In our real time implemention using acoustic sensor at forest we are facing the following external factors affecting the sound propagation at forest[16]-18]. They are listed as follows,
1. Attenuation from geometrical divergence Distance between source and sensor (near & far away)
2. Wind changing the speed of sound profile
3. Temperature changing the speed of sound profile
4. Reverberation inside the forest(Echo)
5. Types of source-Point, Enlarge and Medium objects
6. Humidity affecting the speed of sound profile
7. Diffraction by Barriers (trees and vegetation) and other considerations(reflection,scattering0 etc.,
DATA FOR RESEARCH
The following figure show the acoustic data collected from mudumalai hills. The parameters for elephant sound recordings are listed as follows.
TARGET CATEGORY : 1)ESCALATION -WALKING
TOTAL NO. OF TARGETS : 3
WIND SPEED : 5 miles/hr
WIND DIR : EAST
TEMP : MAX-72 F, MIN-22 C
RELATIVE HUMIDITY : 74%
HEIGHT ABOVE GROUND : 12fT
RECORDING SOUND TIME AND DATE : 10 AM , 10/12/2012
RECORDING LOCATION : MUDUMALAI FOREST
LATITUDE :11.57
LONGITUDE : 76.59
SAMPLING FREQUENCY : 500 micro sec
ELEPHANT TYPE : BABY/ADOLESCENT/ ADULT HOOD
FILE NAME : *3.WAV
VISIBILITY : 10KM
CLOUD COVERAGE : 5%

RESULTS AND DISCUSSIONS

Our proposed results show that due to some external factors the direction of arrival estimation or source localization error are affected. Due to several external factors at forest the source localization is affected by the following reasons. (i) humidity and temperature (ii) enlarged object bear by sensor (iii) reverberation inside the forest(echo). Due to diffraction by barriers,trees and vegetation when the object emit the sound when the wind speed is about 5 miles/hour with in sensor and source with in nearby (25ft) sensor field the source localization achieves high accuracy. Otherwise source and sensor far away from sensor field, temperature variation and reverberation in forest (echo) affects the localization error.

CONCLUSION AND FUTURE SCOPE

This paper, we proposed a methodology for estimating direction of arrival and error analysis for outdoor atmospheric conditions using hyperbolic partial circular array. Our DOA estimation results are very encouraging about using limited battery power and minimum number of sensor nodes. In future our work enhances in two directions (i) identifying more external factors (type of ground surface, thunder, lightning, under water) affecting direction of arrival. (ii) Simultaneously monitor and estimating the direction of arrival for more than one source.

Figures at a glance

Figure 1 Figure 2 Figure 3 Figure 4 Figure 5
Figure 1 Figure 2 Figure 3 Figure 4 Figure 5
Figure 1 Figure 2 Figure 3 Figure 4
Figure 6 Figure 7 Figure 8 Figure 9

References