Forum Replies Created
-
Thank you for providing your network file.
We have found the core issue of the problem.
HARK 2.5 only supported single thread processing, HARK 3.0 on the other hand now uses blas/openblas which allows multi threading.
The workload is spread over all the available CPU which caused the load you were experiencing.
Depending on the kind of process, the threads may not be efficiently used.To reproduce the same load you had using HARK 2.5, please limit the process to a single thread.
This can be done by running the following command.
export OMP_NUM_THREADS=1
export OPENBLAS_NUM_THREADS=1Best regards,
c.santiagoI do apologize for not being more specific.
The network is what we referrer to as the file created and saved using hark_designer.
We have conducted a few test and have not seen a major increase on the load on our CPU when using HARK 3.0 installed on Ubuntu 16.04. For the tests, the network used localization and separation nodes, to be more specific LocalizeMUSIC and GHDSS.
We have however, found that the CPU load was increased significantly(up to 60% in our tests) when installing HARK 3.0 on Ubuntu 18.04.
The source code is the same for both Ubuntu 16.04 and Ubuntu 18.04 so, the reason for the load increase may lie somewhere else.As you mentioned, you are currently using Ubuntu 16.04 so the sharp increase in CPU load should not have effected your environment.
To get to the bottom of the increase in CPU load you have noted, we would like to see the parameters and nodes used in your network file to see if we can replicate your results.
Best regards,
c.santiagoWe are happy to see that everything worked out for you.
It is not ideal for any noise to be present when recording the TSP signal. If the recording contains noise, it will reduce the performance of both sound separation and localization.
Based on the information and example you have provided, it seems you might be having some electrical noise issues.
Although we do not know the core cause of the problem, we can provide you with some possible solutions.
If the noise can be heard through the speaker, please try the following solutions:
1. If the TSP response is being played through a PC, use a device that is less susceptible to noise, such as a USB audio device.
2. If you played the audio using an amplifier, try using another one.
3. In both cases the volume should be set between 20% and 80%. The reason is that distortion tends to occur with characteristics near the limit value of the amplifier.
4. If you are supplying power from an outlet, try changing it to battery-powered.If the noise can only be heard after the TSP signal has been recorded, please try the following solutions:
1. Make sure that the microphone cables are not close to a power supply or related equipment such as AC adapter and/or radio communication equipment (radio).
2. Please check whether the wiring of the effected microphones are loose.
3. Please check whether the wiring of the effected microphones are not entwined with other wires.
4. If you have spare microphones, please try replacing the effected microphones.
5. Please change the USB port where RASP-LC is connected.Thank you for taking an interest in hark.
1a) Do we have to run the sound from the TSP ourselves?
Although synchronized recording can be accomplished using RASP-LC, due to the lack of documentation on how to do it, we recommend using un-synchronized recording for this device. This means the TSP response must be played and recorded by hand.
If you have a digital amplifier unit, connect the 2.5 mm jack on the side of RASP-LC to the digital-in (digital coaxial input Terminal) of the digital amplifier, then connect the output of the amplifier to the speaker. In this case, the quality will be better than playing the TSP audio file from the PC.1b) Files for each source have a different initial indent, how to deal with it?
Because the TSP response will be recorded using un-synchronized recording method, we highly recommend you use our new algorithm Mouth TSP. The initial indents will need to be trimmed off so that all the “sources” align. They do not have to match completely but, the more aligned they are the better.
2) How many pulses do you need to reproduce? 17 or 20?
If you were to use our old method of creating the Transfer Function, at least 16 TSP responses were recommended although using more than that was also okay. With the introduction of our new Mouth TSP algorithm, the amount of TSP response needed has drastically been reduced. A total of about 3-4 responses should be enough. If your current “sources” already contain either 17 or 20 TSP response recordings, they can be used as is.
3) What is the accuracy of the installation of the microphone array and sound source is required for correct operation? 0.02 m will do?
We recommend that the speaker be placed within a 1 meter radius from the microphone array. Please note that the speaker should face the microphone array leaving only a 2-degree margin of error.
4) What are the settings to use in the Harktool5_gui? Sync. Add Num, TSP Offset, TF Calculation Type (use mouth TSP responses or use TSP responses) if responses file as attachments.
Because the TSP response was recorded using un-synchronized method, please follow the steps portrayed in TAMAGO`s transfer function generation manual. As shown in Figure 31 in said manual, most of the values will be left to their default settings.
We recommend that you trim all your “sources” as close to 0 as possible. Use Mouth TSP Record’s Mouth TSP Start and Mouth TSP End by default is set to 16,484 and 32,768 respectively. It is important that the range in frequencies that the sound source is expected to be in is within this period. Since a TSP response contains all the frequencies, the transfer function will be able to detect all types of sources like an alarm clock or a click of a pen. Please note that Mouth TSP Head Margin should be set to 128.
5) What view should the TF have when displayed in harhtool5_gui? (for 8 circle placed microphones(radius 0.09m) and 12 circle placed sources (radius 1 m))
In the case of geometric calculation, in order to generate an impulse from the relative position of the sound source and the microphone, the coordinate information (position information) of the arrangement file of the microphone must be described as accurately as possible. On the other hand, in the case of TSP recording, since impulses are generated from the recorded TSP, the arrangement file of the microphone only needs to have the same number of microphones as the number of channels of the microphone array. Therefore, in the case of TSP recording, there is no problem even if your microphone array and microphone arrangement are different when plotting the microphone placement file.
To check whether the Transfer Function was created correctly, please use HARKTOOL5’s visualize option selecting Localization, Time domain, and Amplitude accordingly and you should see a something resembling a waveform with an initial spike diminishing over time. Please see the attachments bellow.
To answer your original question, if you select a circle as the template in Create Sound Source Positions within HARKTOOL5, set to radius to 1 m, and used a 12 sound source layout file, a circular formation should be displayed.
Attachments:
