localization_exercises

This repositoriy will contain python codes important in using multiple hydrophones to localize sound sousrces.

Exercise #1

Problem statement:

A two hydrophone array receives signals that are recorded in a stereo wave file.

The array is specified thusly:

Separation between the two hydrophones: 1.1 m

The array is oriented with the L channel -> Right channel axis rotated in a horizontal plane 60 deg clockwise from North.

The speed of sound is: 1485 m/s

The question:

For each of the 4 calls in the wav and lable files, what are the likely bearing angles relative to North for the signals received in each call.

Exercise #2

Problem statement:

A binaural array has been deployed at an x,y,z location relative to a local geographic waypoint with the axis from the L channel to the R channel set to a given bearing angle measured positive when clockwise from North.

A variety of calls come from various nearby locations. These locations are unknown. The calls are recorded in a wav format file.

The question:

What can you determine about these calls and how sure are you in your conclusions?

Since this is a model problem, we know where the calls were when the wav file was synthesized. Compare your location notions with the known sources.

An example: (The two hyperbolas are drawn with correlation lags one less and one more than the one computed from the wav file.)

meta data:

generated_sound_high_s2n.wav, generated_sound_low_s2n.wav, generated_labels.txt

array_location = np.array([-20, 0, -10, 340], dtype=np.float64) # x, y, z in meters, bearing angle

local_ref_lat_lon = [48.55841, -123.17327]

speed_of_sound = 1485 # m/s

hydrophone_separation = 1.6 # Fixed separation in meters

when needed, ask for the original source locations

Exercise #3

Problem statement:

Two binaural arrays are deployed, one at Orcasound Lab and one at Sunset Bay. Place some sources at locations out in front of these two arrays.

Model the signal paths to each hydrophone as straight lines

Determine the time of arrival differences for each signal to each of the two arrays.

Convert these to lags in samples and synthesize a stereo file for this lag.

Concantenate all these signals into a wav file with labels for the time at the center of each call.

Now analyze this wav and label file and evaluate the 'accuracy' of localizing each source.

Treat this as a two dimensional problem.

I, Val, am working on a solution but am off today, in the drizzle, to put some lightbulb signals into this problem in the real world. :--)