Malek Lamari
Member level 4
Hi,
I am doing a project in the field of Acoustic Source Localization.
The first part of the project is the simulation of a real environment, for that purpose a Room Inpulse Response (RIR) model must be generated in order to simulate the behavior of the propagated acoustic signal in the room (RIR: include multipath effects, noise, interference ..).
the scenario is as fellow: (see scen.jpg)
* tow or more acoustic receivers (mics) are located at a fixed positions (known).
* one movable source (speaker) that emits the acoustic signal, at each point we calculate a model of the channel between the actual source and each receiver.
* at each point the signal at the receiver is simply the signal at the origin convolved with the impulse response of the channel.
by this manner, we get for every receiver the received signal position dependent characteristics.
at this point, the work starts, witch is applying estimation algorithm in order to extract the needed information for calculating time difference of arrival (TDOA) between receivers.
after that a simple (a bit complicated in fact) mathematical algorithm is needed to estimate the position of the source based on the measured TDOA.
one of the most famous algorithms for TDOA extraction is the Cross Correlation that estimates the resemblance of signals at different time shift.
Cross Correlation gives different amplitudes based of the degree of resemblance of each time deference, and the most high output correspond to the time shift at witch the most significant resemblance occurs.
I implemented the project in MatLab, and successfully get the desired signal located at the receivers. but when applying the Cross Correlation algorithm i get with a strange results. I the output of the xcorr function are not meaningful.
and i cant extract true results from. but if i replace the xcorr function with the "FindDedlay" of MatLab, wich in fact is dedicated to measure the time difference based on CrossCorr, i get some realistic results.
for your convenience i attached the acoustic signals witch are sampled at 22050 Hz at each one of the three mics.
please try to apply your custom CrossCorr algs so that we get with some meaningful results.
I well be very thankful.
Attachments in the zip file:
- source.mat ==> 5 seconds acoustic signal at source sampled at 22050 Hz.
- mic_3.mat ==> 3 columns array contain the signals received at each mic.
mic_3
,1) signal at mic1 ; mic_3,2) signal at mic2 ...
- source.jpg ==> plot of the original signal (source).
- mic_1.jpg ==> plot of the signal received at the mic n1.
- mic_2.jpg ==> plot of the signal received at the mic n2.
- mic_3.jpg ==> plot of the signal received at the mic n3.
View attachment files.zip
I am doing a project in the field of Acoustic Source Localization.
The first part of the project is the simulation of a real environment, for that purpose a Room Inpulse Response (RIR) model must be generated in order to simulate the behavior of the propagated acoustic signal in the room (RIR: include multipath effects, noise, interference ..).
the scenario is as fellow: (see scen.jpg)
* tow or more acoustic receivers (mics) are located at a fixed positions (known).
* one movable source (speaker) that emits the acoustic signal, at each point we calculate a model of the channel between the actual source and each receiver.
* at each point the signal at the receiver is simply the signal at the origin convolved with the impulse response of the channel.
by this manner, we get for every receiver the received signal position dependent characteristics.
at this point, the work starts, witch is applying estimation algorithm in order to extract the needed information for calculating time difference of arrival (TDOA) between receivers.
after that a simple (a bit complicated in fact) mathematical algorithm is needed to estimate the position of the source based on the measured TDOA.
one of the most famous algorithms for TDOA extraction is the Cross Correlation that estimates the resemblance of signals at different time shift.
Cross Correlation gives different amplitudes based of the degree of resemblance of each time deference, and the most high output correspond to the time shift at witch the most significant resemblance occurs.
I implemented the project in MatLab, and successfully get the desired signal located at the receivers. but when applying the Cross Correlation algorithm i get with a strange results. I the output of the xcorr function are not meaningful.
and i cant extract true results from. but if i replace the xcorr function with the "FindDedlay" of MatLab, wich in fact is dedicated to measure the time difference based on CrossCorr, i get some realistic results.
for your convenience i attached the acoustic signals witch are sampled at 22050 Hz at each one of the three mics.
please try to apply your custom CrossCorr algs so that we get with some meaningful results.
I well be very thankful.
Attachments in the zip file:
- source.mat ==> 5 seconds acoustic signal at source sampled at 22050 Hz.
- mic_3.mat ==> 3 columns array contain the signals received at each mic.
mic_3
,1) signal at mic1 ; mic_3,2) signal at mic2 ...- source.jpg ==> plot of the original signal (source).
- mic_1.jpg ==> plot of the signal received at the mic n1.
- mic_2.jpg ==> plot of the signal received at the mic n2.
- mic_3.jpg ==> plot of the signal received at the mic n3.
View attachment files.zip
