Limits...
Object localization using a biosonar beam: how opening your mouth improves localization.

Arditi G, Weiss AJ, Yovel Y - R Soc Open Sci (2015)

Bottom Line: The biosonar sound beam of a bat is directional, spreading different frequencies into different directions.Here, we analyse mathematically the spatial information that is provided by the beam and could be used to improve sound localization.We hypothesize how bats could improve sound localization by altering their echolocation signal design or by increasing their mouth gape (the size of the sound emitter) as they, indeed, do in nature.

View Article: PubMed Central - PubMed

Affiliation: School of Electrical Engineering, Faculty of Engineering , Tel Aviv University , Tel Aviv , Israel ; Department of Zoology , Tel Aviv University , Tel Aviv , Israel.

ABSTRACT
Determining the location of a sound source is crucial for survival. Both predators and prey usually produce sound while moving, revealing valuable information about their presence and location. Animals have thus evolved morphological and neural adaptations allowing precise sound localization. Mammals rely on the temporal and amplitude differences between the sound signals arriving at their two ears, as well as on the spectral cues available in the signal arriving at a single ear to localize a sound source. Most mammals rely on passive hearing and are thus limited by the acoustic characteristics of the emitted sound. Echolocating bats emit sound to perceive their environment. They can, therefore, affect the frequency spectrum of the echoes they must localize. The biosonar sound beam of a bat is directional, spreading different frequencies into different directions. Here, we analyse mathematically the spatial information that is provided by the beam and could be used to improve sound localization. We hypothesize how bats could improve sound localization by altering their echolocation signal design or by increasing their mouth gape (the size of the sound emitter) as they, indeed, do in nature. Finally, we also reveal a trade-off according to which increasing the echolocation signal's frequency improves the accuracy of sound localization but might result in undesired large localization errors under low signal-to-noise ratio conditions.

No MeSH data available.


Related in: MedlinePlus

Angle estimation using the emitted beam. All panels were calculated using an FM signal of 130–40 kHz and a 6.3 mm radius aperture mimicking the signal and mouth gape of M. emarginatus. (a) The correlation function for an object at 25 degrees. The correlation of the reflected echo spectrum with the expected spectra of different angles yields a maximum at the right angle—25 degrees. Note that there is a side lobe (a potential error) at ca 50 degrees. (b) Two-dimensional correlation map between the spectrum of the echo and the beam's spectrum. Hot colours depict high correlation. Note that the solution is circular symmetric—assuming that the range was estimated by the bat based on the pulse–echo delay. If for instance we assume that the azimuth of the object was determined via ITD, as 20 degrees, then only two symmetric solutions (above and below the horizon) are possible (white asterisks). (c) The same as in (a) but for three different objects located at angles 5, 25 and 65 degrees. Note how the main lobe at different angles varies in width and how side lobes appear for certain angles. (d) Left: angular accuracy—the width of the main lobe of the correlation function (see a) for different angles when using the full spectrum (black) or the gamma-tone filter (red). Right: angular ambiguity—the peak to side-lobe ratio for different angles when using the full spectrum (black) or the gamma-tone filter (red).
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4555857&req=5

RSOS150225F2: Angle estimation using the emitted beam. All panels were calculated using an FM signal of 130–40 kHz and a 6.3 mm radius aperture mimicking the signal and mouth gape of M. emarginatus. (a) The correlation function for an object at 25 degrees. The correlation of the reflected echo spectrum with the expected spectra of different angles yields a maximum at the right angle—25 degrees. Note that there is a side lobe (a potential error) at ca 50 degrees. (b) Two-dimensional correlation map between the spectrum of the echo and the beam's spectrum. Hot colours depict high correlation. Note that the solution is circular symmetric—assuming that the range was estimated by the bat based on the pulse–echo delay. If for instance we assume that the azimuth of the object was determined via ITD, as 20 degrees, then only two symmetric solutions (above and below the horizon) are possible (white asterisks). (c) The same as in (a) but for three different objects located at angles 5, 25 and 65 degrees. Note how the main lobe at different angles varies in width and how side lobes appear for certain angles. (d) Left: angular accuracy—the width of the main lobe of the correlation function (see a) for different angles when using the full spectrum (black) or the gamma-tone filter (red). Right: angular ambiguity—the peak to side-lobe ratio for different angles when using the full spectrum (black) or the gamma-tone filter (red).

Mentions: We started off examining the spatial information provided by the wideband Myotis-like signal (mimicking a Myotis emarginatus signal). We simulated the beam of a linear FM down sweep ranging between 130 and 40 kHz [28] with an appropriate M. emarginatus mouth gape (6.3 mm, estimated based on fig. 4 in [13]). We made no assumptions about the bat's HRTF or about the object's distance and frequency response (see Methods). This analysis revealed that the bat's echolocation beam alone provides vast spatial information about the position of an object. We estimated the localization performance using the angular correlation function. This function summarizes the correlation between the actual received spectrum of an echo and the expected spectrum for all angles (figure 2a). The peak of the angular correlation function depicts the angle that is most likely to be the angle of the object. Assuming that the range of the object was estimated using the pulse–echo time delay, and assuming a symmetric beam (as suggested by the piston model and by data [27]), there will always be circular ambiguity, when estimating the position of an object based on the spectral information conveyed by the emitted beam (see the ring of maximum correlation in figure 2b). Under natural conditions, this ambiguity could be solved by using additional cues such as the temporal or spectral information available in the HRTF. For instance, if the bat estimates azimuth based on interaural level or time differences (ILD or ITD, as is often assumed for mammals), then the circular ambiguity converges to only two possible elevations (the intersection of two circles, see white asterisks in figure 2b). These two possible solutions could then be distinguished between based on additional monaural spectral cues. Hence, beam-based spatial information should be thought of as additional spatial information (additional to that provided by the HRTF and the ILD/ITD). Alternatively, a moving bat could analyse two consecutive echoes from different angles and estimate the intersection of two circles resulting in two possible positions. The bat could even use three consecutive echoes to eliminate this dual ambiguity and remain with a single point. This means that, in theory, position could be estimated based on the spatial information conveyed by the emitted beam only.Figure 2.


Object localization using a biosonar beam: how opening your mouth improves localization.

Arditi G, Weiss AJ, Yovel Y - R Soc Open Sci (2015)

Angle estimation using the emitted beam. All panels were calculated using an FM signal of 130–40 kHz and a 6.3 mm radius aperture mimicking the signal and mouth gape of M. emarginatus. (a) The correlation function for an object at 25 degrees. The correlation of the reflected echo spectrum with the expected spectra of different angles yields a maximum at the right angle—25 degrees. Note that there is a side lobe (a potential error) at ca 50 degrees. (b) Two-dimensional correlation map between the spectrum of the echo and the beam's spectrum. Hot colours depict high correlation. Note that the solution is circular symmetric—assuming that the range was estimated by the bat based on the pulse–echo delay. If for instance we assume that the azimuth of the object was determined via ITD, as 20 degrees, then only two symmetric solutions (above and below the horizon) are possible (white asterisks). (c) The same as in (a) but for three different objects located at angles 5, 25 and 65 degrees. Note how the main lobe at different angles varies in width and how side lobes appear for certain angles. (d) Left: angular accuracy—the width of the main lobe of the correlation function (see a) for different angles when using the full spectrum (black) or the gamma-tone filter (red). Right: angular ambiguity—the peak to side-lobe ratio for different angles when using the full spectrum (black) or the gamma-tone filter (red).
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4555857&req=5

RSOS150225F2: Angle estimation using the emitted beam. All panels were calculated using an FM signal of 130–40 kHz and a 6.3 mm radius aperture mimicking the signal and mouth gape of M. emarginatus. (a) The correlation function for an object at 25 degrees. The correlation of the reflected echo spectrum with the expected spectra of different angles yields a maximum at the right angle—25 degrees. Note that there is a side lobe (a potential error) at ca 50 degrees. (b) Two-dimensional correlation map between the spectrum of the echo and the beam's spectrum. Hot colours depict high correlation. Note that the solution is circular symmetric—assuming that the range was estimated by the bat based on the pulse–echo delay. If for instance we assume that the azimuth of the object was determined via ITD, as 20 degrees, then only two symmetric solutions (above and below the horizon) are possible (white asterisks). (c) The same as in (a) but for three different objects located at angles 5, 25 and 65 degrees. Note how the main lobe at different angles varies in width and how side lobes appear for certain angles. (d) Left: angular accuracy—the width of the main lobe of the correlation function (see a) for different angles when using the full spectrum (black) or the gamma-tone filter (red). Right: angular ambiguity—the peak to side-lobe ratio for different angles when using the full spectrum (black) or the gamma-tone filter (red).
Mentions: We started off examining the spatial information provided by the wideband Myotis-like signal (mimicking a Myotis emarginatus signal). We simulated the beam of a linear FM down sweep ranging between 130 and 40 kHz [28] with an appropriate M. emarginatus mouth gape (6.3 mm, estimated based on fig. 4 in [13]). We made no assumptions about the bat's HRTF or about the object's distance and frequency response (see Methods). This analysis revealed that the bat's echolocation beam alone provides vast spatial information about the position of an object. We estimated the localization performance using the angular correlation function. This function summarizes the correlation between the actual received spectrum of an echo and the expected spectrum for all angles (figure 2a). The peak of the angular correlation function depicts the angle that is most likely to be the angle of the object. Assuming that the range of the object was estimated using the pulse–echo time delay, and assuming a symmetric beam (as suggested by the piston model and by data [27]), there will always be circular ambiguity, when estimating the position of an object based on the spectral information conveyed by the emitted beam (see the ring of maximum correlation in figure 2b). Under natural conditions, this ambiguity could be solved by using additional cues such as the temporal or spectral information available in the HRTF. For instance, if the bat estimates azimuth based on interaural level or time differences (ILD or ITD, as is often assumed for mammals), then the circular ambiguity converges to only two possible elevations (the intersection of two circles, see white asterisks in figure 2b). These two possible solutions could then be distinguished between based on additional monaural spectral cues. Hence, beam-based spatial information should be thought of as additional spatial information (additional to that provided by the HRTF and the ILD/ITD). Alternatively, a moving bat could analyse two consecutive echoes from different angles and estimate the intersection of two circles resulting in two possible positions. The bat could even use three consecutive echoes to eliminate this dual ambiguity and remain with a single point. This means that, in theory, position could be estimated based on the spatial information conveyed by the emitted beam only.Figure 2.

Bottom Line: The biosonar sound beam of a bat is directional, spreading different frequencies into different directions.Here, we analyse mathematically the spatial information that is provided by the beam and could be used to improve sound localization.We hypothesize how bats could improve sound localization by altering their echolocation signal design or by increasing their mouth gape (the size of the sound emitter) as they, indeed, do in nature.

View Article: PubMed Central - PubMed

Affiliation: School of Electrical Engineering, Faculty of Engineering , Tel Aviv University , Tel Aviv , Israel ; Department of Zoology , Tel Aviv University , Tel Aviv , Israel.

ABSTRACT
Determining the location of a sound source is crucial for survival. Both predators and prey usually produce sound while moving, revealing valuable information about their presence and location. Animals have thus evolved morphological and neural adaptations allowing precise sound localization. Mammals rely on the temporal and amplitude differences between the sound signals arriving at their two ears, as well as on the spectral cues available in the signal arriving at a single ear to localize a sound source. Most mammals rely on passive hearing and are thus limited by the acoustic characteristics of the emitted sound. Echolocating bats emit sound to perceive their environment. They can, therefore, affect the frequency spectrum of the echoes they must localize. The biosonar sound beam of a bat is directional, spreading different frequencies into different directions. Here, we analyse mathematically the spatial information that is provided by the beam and could be used to improve sound localization. We hypothesize how bats could improve sound localization by altering their echolocation signal design or by increasing their mouth gape (the size of the sound emitter) as they, indeed, do in nature. Finally, we also reveal a trade-off according to which increasing the echolocation signal's frequency improves the accuracy of sound localization but might result in undesired large localization errors under low signal-to-noise ratio conditions.

No MeSH data available.


Related in: MedlinePlus