banner image
No data available.
Please log in to see this content.
You have no subscription access to this content.
No metrics data to plot.
The attempt to load metrics for this article has failed.
The attempt to plot a graph for these metrics has failed.
Extraction of pulse repetition intervals from sperm whale click trains for ocean acoustic data mining
Rent this article for


Image of FIG. 1.
FIG. 1.

Overview of the pulse detection and PRI estimation modules. The input is an acoustic data segment. For each detected pulse, an estimator of the pulse repetition interval PRI and a measure of reliability RM are returned.

Image of FIG. 2.
FIG. 2.

Spectrum Xm (gray curve, frequency resolution 49 Hz) obtained from a short region of 21 ms centered on a sperm whale click. The averaged spectrum Gk is superimposed with a thicker black curve (frequency resolution 938 Hz). A spectrum of the ambient noise is shown with a dotted line.

Image of FIG. 3.
FIG. 3.

PRI and RM plotted against each other. Each dot represents a short sequence of pulses and not a single pulse. Results obtained with full algorithm. Horizontal lines are drawn at PRI = 0.5 and 2. Data are plotted on two subplots according to the manually obtained information on the content of the files.

Image of FIG. 4.
FIG. 4.

Same as Fig. 3 except that results were obtained with the version of the algorithm without Step 3.

Image of FIG. 5.
FIG. 5.

Example of pulse sequences (subset D) returned by Step 4 of the PRI estimation module. The analyzed segment contains a sperm whale click train interleaved with its reflections. A spectrogram of the acoustic data is plotted on the bottom. Crosses represent the output of the pulse detector (ti ). On the upper half of the figure, horizontally aligned white symbols represent automatically identified pulse sequences (subset D). For each sequence, the focal pulse is marked with a diamond, pulses associated to it are marked with circles, and PRI is annotated. Only pulse sequences with RMn  > 4.5 are plotted. A click train was manually extracted, and its clicks are annotated with short bars above the spectrogram.

Image of FIG. 6.
FIG. 6.

Example of pulse sequences (subset D) returned by Step 4 of the PRI estimation module. The analyzed segment contains three or four interleaved sperm whale click trains. See Fig. 5 for an explanation of the symbols.

Image of FIG. 7.
FIG. 7.

Automated identification of sperm whale click trains from a representative dataset. Each dot represents a 5-min file. DVfile is plotted vertically, and the number of detected pulses per minute is plotted horizontally (to quantify the overall pulse density of each file). Upper subplot: Gray dots are weak-SNR SWCT and black dots are medium and high-SNR SWCT (SNR levels are coded as weak = 1, medium = 2, high = 3). Lower subplot: gray diamonds are OTHP alone and black diamonds are SPPT (possible mixed with OTHP). A horizontal line at DVfile  = 5 is given to illustrate a possible threshold for automatically assigning the files to discrete classes (SWCT vs rest).


Generic image for table

Values of parameters used in the PRI estimation module and explanation of their choice. We have observed that in general spectral properties and PRI of a sperm whale click train remain nearly constant for at least 10 consecutive clicks. Accordingly, several parameters were chosen such that PRI and RM are extracted from approximately 10 pulses.

Generic image for table

Overview of dataset. Files of 5-min duration were recorded at the NEMO-ONDE observatory. The files were manually scrutinized and tagged according to their acoustic content. See Sec. III A for an explanation of the tags.

Generic image for table

Search of the threshold Udiss leading to the best identification performance (perfAUC see Sec. III C ). For each value of Udiss , several values of RT were assessed, but in this table only pairs (RT, perfAUC ) where perfAUC is maximal are given.


Article metrics loading...


Full text loading...

This is a required field
Please enter a valid email address
752b84549af89a08dbdd7fdb8b9568b5 journal.articlezxybnytfddd
Scitation: Extraction of pulse repetition intervals from sperm whale click trains for ocean acoustic data mining