Reconstructing the Acoustic Signal of a Sound Source: What Did the Bat Say?

Abstract

Abstract. When attempting to model and understand bat biosonar behaviour, it would be very useful to know exactly what calls the bat emits, that is, what it really says, in the course of its exploration of the world. Calls could be recorded by miniature radio microphone, but such systems are complex and not all bats are sufficiently strong to carry one. In this paper we describe a technique for reconstructing the actual emitted signal of a bat using recordings collected by an array of remote microphones. The theory of the technique is described, experimental results with laboratory-recorded data (for which ground truth is available) are presented, and the performance of the method is discussed.

Reconstruction of the signal produced by a directional sound source from remote multi-microphone recordings

“…In Fig. 4(A), absolute values are within 0.2 and the extreme value is reached only by isolated peaks, as the function normally assumes much smaller values: with respect to the same difference reported in a previous paper 6 and referred to the same orientation, Fig. 4(A) clearly shows how the difference is much smaller at low frequencies.…”

“…4(C) several significant notches are present, like the reconstruction previously reported 6 for the same orientation; with regards to that reconstruction, it was argued that these notches depend on the wrong choice of the samples in the microphone recordings corresponding to when the original signal is detected. 6 This guess was true in the reconstruction for (20 , À 20 ), when sample indices were chosen analytical calculation (that turned out to be inaccurate); 6 but in this case there are only 2 recordings contributing to the reconstruction, and their timing relationship is correct.…”

“…Estimation of source orientation has already been investigated by the authors 3-5 as well as signal reconstruction. 6 The contribution of this paper is in using strategies for precisely determining, in an automatic way, time shifts between the beginning of each microphone recording. The first strategy uses a matched filter on each recording, based on an estimation of the same recording computed using other microphones' signals.…”

“…As a first attempt, the first sample of the signal from the transducer in each microphone recording was calculated by considering the distance of each microphone from the transducer. 6 Such a calculation is not always accurate because microphone positions are not exactly measured and ultimately it was necessary to manually select the sample at which the original signal was detected by microphones in order to get the actual recorded signals. 6 In the following, two strategies for accurately calculating the time shifts Dt m are investigated: details are in Secs.…”

“…Estimation of source orientation is a precondition of reconstruction that has been investigated by the authors in previous work. [3][4][5] Afterward, the method for reconstructing the signal produced by the transducer was presented 6 and a strong dependence of method's performances on the arrival times when microphones detect the signal from the source was pointed out. Therefore, the work described in this paper, besides providing a full description of the method for reconstructing the signal produced by a source, investigates two strategies for automatically determining the signal arrival time in microphone recordings, that is, microphone time shifts, that are used as inputs of the method.…”