Detection of the carrier's audible announcement

Hello, I need help with my school work. The aim of the work is to detect the sound cues that accompany the sound announcements of public transport stops. The audio track is pre-recorded and inserted into the script, so it is not a real-time audio track. The final form should either show a graph of the audio track with a point showing the start of the announcement. I've tried to address this in a few ways, but this is the closest I've come: I used the voiceActivityDetector function, which worked for the most part, but! I found that the ambient sounds largely interfered with this detection. Despite filtering the different frequencies using windows (or using the lower pass), the result still looked the same.

So my question is: Do you know how to filter out ambient noise or if there is a better function than voiceActivityDetector?

Thank you very much for the answer and sorry for my bad English.

Answer 1

[William Rose]

@Matyas Rychetsky,

I do not know of anything better than the voiceActivityDetector.

The examples here are very good, in my opinion. The examples include plots of "probability of speech" versus time. The VAD returns a vector with probability of speech at each time. You can specify a probability threshold for detection.

If it still is not working, you could try a bandpass filter, such as 300-3000 Hz, the old "C-band" for voice, and see if that helps.

Your English is excellent!