Hello! While setting up a network in HARK_DESIGNER, we’ve faced a few difficulties:
1) I can place the SemiBlindICA module on MAIN and MAIN_LOOP.
In the first case (MAIN), the system works without delay. But I am not sure that the processing takes place in the right way, as you wrote earlier.
In the second case (MAIN_LOOP), the system works with the accumulation of a pause, that means HARK starts processing data received from microphones in 3 seconds in the first cycle, in 6 seconds – in the second cycle, and so on. I tried to change the settings of the HarkMsgsStramFromRos and SemiBlindICA modules, but did not get any changes. I think that the accumulation of pause occurs in the separation network.
– The LENGTH and WINDOW_LENGTH parameters are in many modules. Do they have to be the same everywhere?
– In my case MSG_BUFFER_NUM and DATA_BUFFER_NUM in StreamFromHarkMsgs are equal in size of the buffer sent in the HarkWave message, is this correct? Or do I need to specify the queue size which is set when writing messages to the topic?
3) The HarkWave message contains count. It is equal to seq from header. Am I doing the right thing?
4) HarkWaveVal from HarkWave implies vector <float>. When saving data directly to a file, we get clear sound. After writing a message to HarkWave topic, we catch it by StreamFromHarkMsgs node and write to the file. In this case strong noises appear in the file. Do I need some kind of additional changes in order to transmit through the topic, or do I need to write to the topic just a set of bytes?
5) When recording in ROS-topic, we set the size of the stored queue. Do I need to Indicate it in some of the nodes in hark_designer?