Analysis of Acoustics
An analytical system has been developed to enable interpretation of observations from the habitat sensors. A focus on the measurement of the soundscape has enabled the habitat sensor-server system to sense multiple environmental variables (images, temperature, light, humidity, etc.). Acoustics has been the focus or our sensing activity. The acoustic signals that stream into the digital library from the habitat sensor servers represent a huge challenge to large scale movement of data over wireless networks to local servers and then to remote regional servers. To approach the analysis of acoustics, both a general and a specific analysis is considered.
General Analytical Approach. The general approach addresses using acoustic signals to assess habitat quality. The general approach partitions an acoustic signal into 1 KHz intervals and computes the power in each of the intervals. An algorithm developed in MATLAB is used to compute the amount of power in each 1 KHz frequency of a sound sample. The analytical process was automated analyze and visualize potentially thousands of acoustic clips sampled at regular intervals over long time periods. An examination of acoustic time series reveals important patterns which characterize a place. Research identified that technophony generally occurs between 1-2 KHz and Biophony generally occurs between 2-7 KHz.
[photopress:Bio_panel_1.jpg,full,centered]
The figure above is a visualization of acoustic signals showing an ocillogram, a spectrogram, a power spectral density plot and a bar chart of power in 1 KHz frequency bins.
A ratio of biophony to technophony is calculated to create a normalized index (-1 to +1) of habitat quality based on the nature of the soundscape.
The figure below is based on an annual half-hourly sample of the soundscape. This daily pattern is recorded from a rural soundscape in Okemos, MI based on 48 observations per day. The acoustic habitat quality index (AHQI) at this site has a generally strong average biological signal (> 1.0) that dips as people commence commuting around 0730 Hrs, increases until 1100 Hrs, decreases steadily until 2000 Hrs and then biophony increases as human activity lessens in the evening. This figure was derived by automatically processing 17,520 acoustic files to yield this annual pattern.
[photopress:AHQI_Daily_Annual_GH.jpg,full,centered]
Species Specific Analytical Approach. Analysis of species specific acoustics is more complex and thus more problematic. However significant advances in identification of species using signature matching have been accomplished. To identify a species we listen to the sound and identify a signature in the image and extract the signature from the spectrogram image of a sound sample. A signature extraction system using MATLAB has been developed to automate the process of signature development. The signatures of species are then matched to sounds sampled from soundscapes sampled using the habitat sensor platforms.
The process for determining the time of day when the spring peeper signals most often during May is described below. The spring peeper signaling is shown in the spectrogram below (left image). The signature of the spring peeper (right image) is extracted from the spectrogram. It is this signature that is searched for in the acoustic samples.
[photopress:spring_peeper_signature.jpg,full,centered]
A result of the signature match for the spring peeper is shown below. The signature is matched against acoustic samples recorded 48 times per day during May 2006 from a pond site near Okemos, MI. The figure below shows the mean (se) match for each half hourly period. The line at 0.38 is a threshold of match based on listening for spring peeper signals in the sound samples. In this example, spring peepers signal after 2100 and cease signaling after 0600 hrs.
[photopress:spring_peeper_search_correlaton.jpg,full,centered]