2

I have a corpus of short speech samples from Kiswahili speakers, and I want to detect the number of syllables in each audio recording. How should I approach this task?

Background: I asked a conceptual/programming variant of this question on StackOverflow showing my attempt at peak detection in R. Users migrated the question to CrossValidated, but someone suggested that SP might be a better home. I asked how to handle the situation on SP Meta, and a user gave me helpful advice to frame this question more conceptually on SP.

Community
  • 1
Eric Green
  • 121
  • 2
  • I assume that is a research topic of its own. I am not a linguistics guy, but I have a feeling that you have to tackle a quite hard problem there, if you intend to design the algorithm on your own. – M529 Apr 15 '16 at 11:36
  • This is a simplification of speech recognition; I'd approach the problem from that direction. [Here](https://www.youtube.com/watch?v=JL4KJznAySE)'s a tutorial. The only problem is that you will need a lot of data, and you don't seem to have it, so your low-tech solution might be right. Try using a different feature such as the [MFCC](https://en.wikipedia.org/wiki/Mel-frequency_cepstrum). – Emre Apr 15 '16 at 22:29

0 Answers0