Вы находитесь на странице: 1из 12

PROJECT REPORT:

AIM :
Analysis of Bat signal using Fourier Transform.

INTRODUCTION:
A general introduction about Bats
Bats generate ultraviolet via the larynx and emit the sound through the nose or the open mouth. Bat calls range in frequency from 14,000 to over 100,000 hertz, well beyond the range of the human ear (typical human hearing range is considered to be from 20 Hz to 20,000 Hz). Bats use echolocation to navigate and forage, often in total darkness. They generally emerge from their roosts in caves, attics, or trees at dusk and hunt for insects into the night. Their use of echolocation allows them to occupy a niche where there are often many insects (that come out at night since there are fewer predators then) and where there is less competition for food, and where there are fewer other species that may prey on the bats themselves. When searching for prey they produce sounds at a low rate (10-20/sec). Individual bat species echolocate within specific frequency ranges that suit their environment and prey types. This has sometimes been used by researchers to identify bats flying in an area simply by recording their calls with ultrasonic recorders known as "bat detectors". However echolocation calls are not always species specific and some bats overlap in the type of calls they use so recordings of echolocation calls cannot be used to identify all bats. In recent years researchers in several countries have developed "bat call libraries" that contain recordings of local bat species that have been identified known as "reference calls" to assist with identification. Some moths have developed a protection against bats. They are able to hear the bat's ultrasounds and flee as soon as they notice these sounds, or stop beating their wings for a period of time to deprive the bat of the characteristic echo signature of moving wings which it may home in on. To counteract this,

the bat may cease producing the ultrasound bursts as it nears its prey, and thus avoid detection.

Terminology used for analysis:Frequency Modulation and Constant Frequency: Echolocation calls can be composed of two different types of frequency structures: frequency modulated (FM) sweeps, and constant frequency (CF) tones. A particular call can consist of one, the other, or both structures. Intensity: Echolocation calls have been measured at intensities anywhere between 60 and 140 decibels. Certain micro bat species can modify their call intensity mid-call, lowering the intensity as they approach objects that reflect sound strongly. Harmonic composition: Calls can be composed of one frequency, or multiple frequencies comprising a harmonic series. In the latter case, the call is usually dominated by a certain harmonic. Call duration: A single echolocation call (a call being a single continuous trace on a sound spectrogram, and a series of calls comprising a sequence or pass) can last anywhere from 0.2 to 100 milliseconds in duration, depending on the stage of prey-catching behaviour that the bat is engaged in. For example, the duration of a call usually decreases when the bat is in the final stages of prey capture this enables the bat to call more rapidly without overlap of call and echo. Reducing duration comes at the cost of having less total sound available for reflecting off objects and being heard by the bat. Pulse interval: The time interval between subsequent echolocation calls (or pulses) determines two aspects of a bat's perception. First, it establishes how quickly the bat's auditory scene information is updated. For example, bats increase the repetition rate of their calls (that is, decrease the pulse interval) as they home in on a target. This allows the bat to get new information regarding the target's location at a faster rate when it needs it most. Secondly, the pulse interval determines the maximum range that bats can detect objects. This is because bats can only keep track of the echoes from one call at a time; as soon

as they make another call they stop listening for echoes from the previously made call. Therefore the bat constantly has to make a choice between getting new information updated quickly and detecting objects far away.

METHODOLOGY
FLOW CHART

For analysis of Bat signal we obtain the signal data from the signal processing information base (SPIB): http://spib.rice.edu/spib.html. The echolocation chirp recorded from a large brown bat (Eptesicus fuscus) is used for analysis. The sampling period is T=7microseconds, and there are N = 400

samples in the data set. The time domain signal oscillates and rises in amplitude. The magnitude spectrum explains a little more. With the help of MATLAB we plot the frequency spectrum of Bat signal as shown in Fig(2). But from frequency spectrum it is difficult to identify the different signal. So we did the spectrogram of signal using Hann window (128) and Boxcar window (128). Now, we plot frequency vs time plot of Bat signal using Hann window and Boxcar window of length 128 as shown in the Fig(3) and Fig(4). From these figures it is easily reveals three descending chirps. A chirp is a signal segment where the frequency rises or falls over time. Strictly speaking, of course, a chirp is not a narrowband signal. But locally, at least, the signal energy is contained in a narrow spectral range. MATLAB code for above process is given below.

MATLAB CODE
1.
% First we load the file. Load bat.txt x=bat; t=7*1e-6*(1:400); % Now plotting bat signal. Subplot (2, 2, 1); plot(t, x); x label (' Time(s) '); y label ('Amplitude'); title ('Time domain'); % plotting fft of bat signal. k=(1:400)*14286/400; subplot (2,2,2); plot (k,abs(fft(x))); xlabel ('Frequency(Hz)'); ylabel ('Amplitude'); title ('Frequency domain'); % now doing spectrogram Fs= 1e6/7; w0=hann(128); subplot(2,2,3); spectrogram(x,w0,120,256,1e6/7); % noverlap=100,nfft=256,Fs=1e6/7; title(' Spectrogram for w0=hann(128),noverlap=120,nfft=256,Fs=1e6/7'); subplot(2,2,4);

w1=boxcar(128); spectrogram(x,w1,120,256,1e6/7); % noverlap=50, nfft=256,Fs=1e6/7; title(' Spectrogram for w1=boxcar(128),noverlap=120,nfft=256,Fs=1e6/7');

2. Code for comparing different window length and overlap for Hann window and Boxcar window in MATLAB. Load bat.txt; x=bat; subplot (2,2,1) spectrogram(x,128,100,512,1e6/7); title ('spectrogram for w=128,noverlap=100,nfft=512,Fs=1e6/7'); subplot(2,2,2) spectrogram(x,128,120,1024,1e6/7); title('spectrogram for w=128,noverlap=120,nfft=1024,Fs=1e6/7'); subplot(2,2,3) spectrogram(x,64,10,512,1e6/7); title('spectrogram for w=64,noverlap=60,nfft=512,Fs=1e6/7'); subplot(2,2,4) spectrogram(x,64,60,1024,1e6/7); title('spectrogram for w=64,noverlap=10,nfft=1024,Fs=1e6/7');

DISCUSSION AND RESULTS


The echolocation chirp recorded from a large brown bat (Eptesicus fuscus) is used for analysis. The sampling period is T=7microseconds, and there are N = 400 samples in the data set. The time domain signal oscillates and rises in amplitude. The magnitude spectrum explains a little more. There are frequencies between 20 kHz and 50 kHz, centred more or less strongly around a spectral peak at some 35 kHz. The spectrum appears to be bimodal. From a Fourier domain perspective, we cannot tell whether the modes are frequency components that appear at certain times, one after the other, or whether they substantially overlap and the bimodality is an artefact of relatively of relatively weaker middle frequencies.

Time representation of Bat Signal

Fig(1)

Frequency representation of Bat Signal

Fig(2)

Our knowledge of the echolocation pulses content changes dramatically when we develop the time-frequency map. The spectrogram is one of the most useful and widely used tools in signal processing. The spectrogram is, like the very first sound spectrograph, the running output of the Fourier transform of a signal, typically defined as the intensity of the short-time Fourier transform (STFT).The classical means to calculate the STFT (and spectrogram) involves windowing a time-series signal into discrete intervals in time, which are each Fourier transformed in turn to obtain the power spectrum of the signal as a function of time. We use a boxcar window of width N = 128 to generate the local spectral values, overlapping successive windows by M = 120 samples. This reveals three descending chirps and shows that the time-frequency plot is at

least tri-modal. The bar-shaped artefacts most visible in the lower frequencies appear to correlate with window alignment. We try a Hann window function of length N and overlap M to improve the local frequency estimates, as shown in Figures. This reduces the time-frequency artefacts, as one might expect. However, Hann windowing has the added benefit of resolving the timefrequency mode of highest initial frequency into two chirps (A chirp is a signal segment where the frequency rises or falls over a time.); the echolocation pulse in fact contains four modes.

Spectrogram of Bat Signal using Hann window

Fig(3)

Spectrogram of Bat Signal using boxcar window

Fig(4)

Spectrogram using different window size

Fig(5)

Conclusion
A method for processing signals based on optimising the fourier transform for Bat signal has been presented. We have demonstrated that our processing scheme offers spectrograms with improved resolution of signals containing multiple chirps. From Fig(1) we can say that the time domain signal oscillates and rises in amplitude. The magnitude spectrum explains a little more. From frequency domain as shown in Fig(2) we observe that there are frequencies between 20 kHz and 50 kHz, centred more or less strongly around a spectral peak at some 35 kHz. The spectrum appears to be bimodal. By looking at the plot of spectrogram as shown in the Fig(3) and Fig(4) it reveals the three descending chirps and shows that the time frequency modal is at least tri-modal. The bar shaped artifacts most visible in the lower frequencies appear to correlate with window alignment. From Fig(3) we saw that local frequency estimates are improved with the help of Hann window. This Hann windowing has the added benefit of resolving the time-frequency mode of highest initial frequency into two chirps. In Fig(5), the effect of different window length and number of overlap is clearly illustrated.

References :-

CAPUS, C. & BROWN K. 2003. Short-time fractional Fourier methods for the time-frequency representation of chirp signals. J. Acoust. Soc. Am. 113 (6), 3253-3263. Detecting non-stationary signals using fractional Fourier methods By A. T. Catherall and D. P. Williams Defence Science and Technology Laboratory, Physical Sciences, Winfrith Technology Centre,Dorchester DT2 8WX. SCALE DOMAIN ANALYSIS OF A BAT SONAR SIGNAL Branko Ristic and Boualem Boashash Signal Processing Research Centre Queensland University of Technology GPO Box 2434, Brisbane, Q 4001, Australia Signal analysis time,frequency scale and structure by Ronald L. Allen and Duncan W. Mills. The analysis of multiple linear chirp signal by Chris Capus Image analysis research group, Heriot Watt university and Yuri Rzhanov ,Laurie Linnett of University of New Hemisphere.

Вам также может понравиться