measuring rate of speech in realtime

问题:

I'm looking for a quick and simple way to measure the rate at which I am speaking in real time. Course grained approaches or approximations are sufficient. The idea is to write a simple app/widget that at least tells you to speed up or slow down while speaking. Measuring things like pitch and volume might also be nice.

I assume this can be done simply with a variety of speech recognition libraries, but I am familiar with none of them and quick glances at the documentation do not give a simple answer or even a good notion of which library to look at in more depth.

Ideally the library would have a python interface, but anything that can run on linux/osx/android would suffice. iOS is not viable. Usability of the library/examples is more important than performance.

回答1:

Efficient speech rate detection can be done without speech recognition because speech recognition requires a lot of computation. The algorithm is described in

Dagen Wang and Shrikanth S. Narayanan. Robust speech rate estimation for spontaneous speech.

Ali Ziaei, Abhijeet Sangwan, John H. L. Hansen. A Speech System for Estimating Daily Word Counts

and uses peak detector for very fast estimation of the word counts.

There is no open source implementation though, but it is worth to implement it yourself.