Rajendar Bahl

Simulation of backscattering of high frequency sound from complex objects and sand sea-bottom

This paper describes the simulation of backscattering of high frequency sound from complex objects and sand sea-bottom. Backscattering data from complex objects and seabottom were generated using a Universal High Resolution Imaging Sonar Simulation Toolkit (UHRISST) developed by the authors. Our approach here involves the approximation of objects and the sea-bottom through a series of facets that are small compared to the wavelength. The Target Strength of standard objects like the sphere, cylinder, and cone predicted by our model have been compared with various high frequency approximation formulations and produced a good fit. For the prediction of the backscattering strength from the sea-bottom, we have taken into account most of the small-scale features that represent the sand grain properties, intermediate features that represent seabottom roughness and large features that block propagation. Sea-bottom backscattering predictions made using this model were compared with high frequency scattering strength values measured off Panama City, FL, USA, and produced a moderate fit. Nonparametric statistical analysis was carried out on the bottom backscattering signals from a flat isotropic sea-bottom. The probability distribution function and probability of false alarm function of the bottom backscattering data was measured and found to largely follow the normal distribution. >

Underwater behavioral study of Ganges river dolphins by using echolocation clicks recorded by 6-hydrophone array system

Targeting the 9 Ganges river dolphins inhabiting the Narora area in Ganges river system, the first recordings using an compact acoustic observation system, which consists of a high frequency 6-hydorophone cross array connected with a PC on land-base by a wireless LAN system have been carried out during 17 to 21 February 2008. The system also uses the prototype of a GPRS modem-based real-time data transmission system to share the processed data with the world over the internet. The array structure consists of a horizontal cross array formed by 2 arms each of 1.6 meter length with a hydrophone on each corner and at the intersection forming two equi-spaced 3-hydrophone linear SBL system for omni-directional horizontal coverage, and one additional hydrophone which is fixed 0.8 meter below the center hydrophone of the cross array also forming an equi-spaced triangular SBL system for vertical measurement. The clicks of the dolphins received on each hydrophone were sampled at a rate of 500 kHz per channel. Total recording was done for approximately 40 hours due to power supply interruptions. However, the GPRS modem-based real-time data transmission system worked well during the recordings. The results of the basic analysis of the data show that the dolphins are found swimming up and down the river. The results are in accordance with the visual observation results. Underwater behavior of the dolphins approaching near to the array, which could not otherwise have been observed by visual survey, has been mainly analyzed in this paper. Based on ICI during this period of approaching to the array, we estimated ldquowhat this dolphin is seeingrdquo or interested in. We invoked the idea of lag time that the dolphin requires to perform echo-location process and to emit the next click to the target [7]. Defining the lag time as DeltaP (pre-click time) and assuming that DeltaP is constant, DeltaP and distance D to the target where the dolphin is looking at is calculated by using ICI and range between the dolphin and the array, which is measured by the array. Utilizing available on-axis clicks from the data of the preliminary experiments in Budhabalanga river, the value of DeltaP is calculated. Constant DeltaP is estimated as 14.3 milliseconds. To determine D of the Narora data, we applied it when the dolphin was approaching to the array within 20 meters. The results show that the dolphin does not look at any object beyond 20 meters range. On the other hand, when ICI reduced below the constant DeltaP (14.3 milliseconds) , it is estimated that the dolphin is in a different stage of the echo-location process.

Localization of sperm whales in a group using clicks received at two separated short baseline arrays.

In this paper, a sperm whale click analysis scheme is proposed in order to calculate the position of individual sperm whales in a group using data received at two arrays deployed near the surface. The proposed method mainly consists of two parts: short baseline (SBL) with classification and long baseline (LBL) with class matching. In SBL with classification, a click is automatically detected, and its direction of arrival is calculated. The clicks are then classified based on their direction vectors. The class data are then sent together with direction data and matched to the other arrays class data. LBL with class matching is used for localization. The classification algorithm can be used to estimate the number of whales clicking and to list potential candidates for LBL matching. As a result, the proposed method is able to localize the positions of the whales in a group. The performance of the proposed method is evaluated using data recorded off Ogasawara islands with two arrays near the surface. The three-dimensional underwater trajectories of six sperm whales are extracted to demonstrate the capability of the proposed method.

Beam pattern estimation of clicks of a free-ranging Ganges river dolphin

The first recordings of bio-sonar clicks of free-ranging Ganges river dolphin (Platanista gangetica) were made in Budhabalanga river in Orissa, India. The system used for recording is a 3.2 meter long array composed of three hydrophones forming an equispaced linear SBL array and another two hydrophones in conjunction with the central hydrophone forming a small SSBL triangular array in a plane perpendicular to the array axis. The array structure was deployed both in horizontal and vertical configurations. The clicks of the dolphin received on each hydrophone were sampled at a rate of 500 kHz per channel. From the sampled clicks data, the sonar transmit beam pattern of the dolphin in both horizontal and vertical plane was estimated using difference of ASL (apparent source level of the click signal) and the absolute angles of the dolphins location between the central hydrophone and the other two peripheral hydrophones for the on-axis clicks. To estimate more accurate beam pattern, conditions for selection of on-axis clicks were well-considered, and thus we propose a new criteria for determining on-axis clicks. In this method, we assume that the dolphins beam pointing direction is towards the central hydrophone and its beam pattern is symmetrical. However, actual measurement of ASL from three hydrophones shows that values at the peripheral hydrophones are not always same. This suggests that the dolphins beam is not pointing exactly towards the central hydrophone so that the central hydrophones ASL is not the real peak value of the beam. We propose a refinement in the beam pattern estimation method using interpolation and curve fitting to obtain an improved estimate of the Ganges river dolphins beam pattern. By analyzing all available clicks, we have estimated the dolphins beam pattern in both horizontal and vertical plane. The 3-dB beamwidth is found to be approximately 10 degrees in horizontal plane and 14 degrees in vertical plane. In this paper, we propose the new beam pattern estimation method using interpolation and curve fitting.

Performance of 4- and 8-state TCM schemes with asymmetric 8-PSK in fading channels

For Rayleigh fading channels, Ricean fading channels with a small parameter and shadowed Rician channels with the code design criteria are to maximize the effective length as well as the minimum product distance of the code. In this paper, we use two different asymmetric 8-PSK signal sets in 4-state and 8-state rate 2/3 trellis-coded modulation (TCM) schemes and show performance gain over TCM schemes with symmetric 8-PSK constellations. The performance gain is due to an increase in the minimum product distance compared to that of the best known TCM schemes of comparable states using symmetric 8-PSK signal sets while the effective length remains same. Simulation is performed over the Rayleigh, Ricean, and shadowed Rician fading channels to demonstrate the performance gain of the asymmetric 8-PSK TCM schemes over the symmetric 8-PSK TCM.

Long duration real-time observation of Irrawaddy dolphins in Chilika lagoon

The authors have developed a passive acoustic observation system to monitor Irrawaddy dolphins (Orcaella brevirostris) in Chilika lagoon. The system is designed to monitor the detailed underwater behavior of dolphins using a Short Base Line (SBL) system all through the day and night. The underwater sounds are recorded at a sampling frequency of 500 kHz with 16 bits resolution. This system was deployed in Chilika lagoon from 29th January to 3rd February and on 18th February in 2007. Experiments were carried out all through the night. By replacing the batteries and hard disks, the observation system was made to keep running all the time. Consequently, the authors succeeded in obtaining click data from Irrawaddy dolphins for a continuous duration of about 21-hour consecutively for each trial and about 1TB of acoustic data was obtained in total. The data contained many high frequency signals, which are considered to be the snapping noise made by shrimp, which are typically found in warm shallow water areas such as Chilika lagoon. A de-noise calculation is developed and applied to the obtained data. A clear picture of trajectories of dolphins in the turbid environment of the Chilika lagoon can be observed that has never before been seen. In this paper, detailed ideas and descriptions of the set up and deployment of the system are presented. Furthermore, the de-noise calculation and the observed dolphin trajectories will be discussed.

Estimation of instantaneous frequencies using iterative empirical mode decomposition

A novel instantaneous frequency-based time–frequency representation is proposed for the analysis of multicomponent signals. The concept of frequency translation is innovatively combined with the empirical mode decomposition algorithm to formulate an iterative procedure, referred to as the iterative empirical mode decomposition, to separate the components present in a signal at a suitably selected frequency resolution. The instantaneous frequency and amplitude estimated on the separated components are used to form the new time–frequency representation. The iterative empirical mode decomposition is assessed for component resolvability, and the performance of the aforementioned time–frequency representation is compared with several other time–frequency representations based on visual inspection and using objective criteria. The Hilbert spectrum formed using the iterative empirical mode decomposition not only provides high concentration of energy about the components’ instantaneous frequencies at high signal-to-noise ratio, but also good resolution while keeping the interference terms at a minimum.

Vocalization based Individual Classification of Humpback Whales using Support Vector Machine

Visual surveys and DNA analysis hold primary importance in marine mammal observation. Especially, recognition of whales and their population density estimation rely heavily on these costly and tedious methods. Acoustic survey of vocalizing cetaceans, in contrast, is emerging as a promising technique for efficient, automatic, non-invasive and convenient observation and analysis of such vocally active organisms. However, it is, yet, too premature to replace the conventional methods altogether. In this paper, we have presented our results on recognition of individual humpback whales based on their vocalization data. Cepstral coefficients from song units extracted from audio records of seven humpback whales, re-sampled at 8KHz, were subjected to cepstral analysis. The extracted coefficients were used to develop multi-class support vector machine (SVM) classifier model. The test phase results indicated classification accuracy as high as 99% as compared to earlier best results of 88%, achieved by Gaussian mixture model (GMM) trained on cepstral coefficients. Furthermore, this improvement was attained with a highly reduced training dataset size. It was also verified that the song duration of approximately 128 seconds was sufficient for reliable identification in test phase for given dataset.

A multimodal detection model of dolphins to estimate abundance validated by field experiments

Abundance estimation of marine mammals requires matching of detection of an animal or a group of animal by two independent means. A multimodal detection model using visual and acoustic cues (surfacing and phonation) that enables abundance estimation of dolphins is proposed. The method does not require a specific time window to match the cues of both means for applying mark-recapture method. The proposed model was evaluated using data obtained in field observations of Ganges River dolphins and Irrawaddy dolphins, as examples of dispersed and condensed distributions of animals, respectively. The acoustic detection probability was approximately 80%, 20% higher than that of visual detection for both species, regardless of the distribution of the animals in present study sites. The abundance estimates of Ganges River dolphins and Irrawaddy dolphins fairly agreed with the numbers reported in previous monitoring studies. The single animal detection probability was smaller than that of larger cluster size, as predicted by the model and confirmed by field data. However, dense groups of Irrawaddy dolphins showed difference in cluster sizes observed by visual and acoustic methods. Lower detection probability of single clusters of this species seemed to be caused by the clumped distribution of this species.

A micropower analog hearing aid on low voltage CMOS digital process

A two-chip analog micropower hearing aid circuit is developed which is based on a low voltage three micron CMOS process. The novel features of the circuit are the use of adaptive biasing of MOS Translinear Loop (MTL) circuit and an innovative application of an adaptive technique in reducing the value of a degenerating linearising resistor in the input differential stage of the AGC block. The above two measures enable reduction of power consumption and external component count. Class-D amplifier provides high conversion efficiency at the output stage. The proposed configuration is now under integration for developing a one chip general purpose CMOS analog hearing aid with capability to operate with 1.0 volt supply voltage.