Calculating Intersection Times with SoundCheck

Find the start and stop times of any recorded signal with SoundCheck’s intersection post-processing step. In this video, we’ll show you how to define a threshold, locate where your waveform crosses it, and calculate timing for applications like wake-word latency or response detection.

Additional Resources for SoundCheck

Detailed explanations of how to use SoundCheck can be found in the SoundCheck manual.

Video Transcript: Calculating Intersection Times with SoundCheck

In this video, I’ll show you how to find the start time of a recording. 

Here I have a sequence which recalls a wave file. I’m using the ITU real speech wave file which is provided with SoundCheck as an example. 

I’ll apply the recall step, and now we have the waveform in the memory list. 

Let’s say I want to find the start time of this waveform, which is around 0.5 seconds. 

We can do this with an intersection post-processing step, but first I need to define a signal level or a threshold. 

We need to select a value that is above the noise floor but within the signal of the recording you’re using. Most of the signal in this waveform lies above 30 m FS, so I’ll use 30 mFS. This means SoundCheck will find the point where the waveform crosses the 30 mFs value. 

Once this value is defined, I use an intersection post-processing step. I select the wave file or the recording as my operand A and the signal level that I have defined as operand B. 

We’ll use the “search up”  option, which starts at the beginning of operand A, which is the recording, and finds the point where it intersects with the signal level. 

“Search down” would do it the other way, and “return all intersection points” will identify all points at which the waveform corresponds to the signal level. 

Now let’s run the sequence, and we get the start time as a value, which is 492 milliseconds, approximately 0.5 seconds as we saw in the waveform graph. 

This is a simple example of how to calculate the start time.  This could be used, for example, for measuring the latency of a response to a wake word. If you wanted to find the stop time, you could use the “search down” function in the intersection step, and you could even calculate waveform duration by subtracting the start time from the stop time.