Author Options:

Determining speech rate from audio Answered

Hi, my name's Oli and I am a speech therapist - aka speech pathologist in US.

I am looking for a way (e.g. an app) that can determine how fast someone is speaking, in realtime. This kind of biofeedback would be really useful for many people with communication disorders where learning to speak slower can help a lot (e.g. stuttering, dysarthria). I can also see it being useful for teaching public speaking to students etc. 

Does anyone know of anything out there (hardware or app) that can do this? Or where I might find someone I could pay to develop this app? 

Many thanks,



what an excellent idea!

my lil brother has always had a speech impediment and has inspired me to want to learn why, how to help him and others like him!

I have taken him to a class before and I think I was about 20 and he was 17

I fell asleep in my chair halfway through the one on one session

this helped me realize how painfully boring these classes were and why he gave up entirely on speech Therapy.

I guarantee if there was "cool" apps/technology to help these kids stay engaged it would definitely help many pp.

THANKS for posting

it inspired me

What about that favourite of musicians everywhere the metronome?


2 years ago

Try a Linux Operating System. For example, Ubuntu.

It includes an application that searches for all the registered programs with the topic of your choice. For example, searching for "speech recognition" appear a list of free installable programs...

"PocketSphinx is a lightweight, cross-platform, speaker-independent,
large-vocabulary continuous speech recognition engine..."

... then another,


And a list of add-ons also installable to complement the main application.

In your case, you should search for a program that has a community and/or add-ons in a language like Perl or Python where there are a big number of people making programs, perhaps just for fun (perhaps).

If you cannot find a program that would just keep counting the produced text with the words that the engine is 'hearing', you could find out the address of the users community site for that software, and ask there more precisely. For example: "... I want to get a constant average of the number of words heard per minute..."

You could even ask that the program could check the nature of every word... If the word exists, if it is a common word, if it is an aggressive term. Or even to keep grouping words related to the same subject, counting the number of time that they repeat, etc. And even, make a sound when ever the number of words per minute exceeds a certain amount.

I hope this helps!

Thanks for this, sorry for my slow reply - I've only just seen this.

Linux sounds promising, I will check it out. It doesn't so much matter whether the words are accurately detected but just that the software detects that a word has been spoken. I'll look into PocketSphinx and Julius.

Thanks again

First you need to define your metrics. Are you talking about words per minute or some other measurement? Though i fail to see how knowing words per minute is going to help with anything. People talk at all different rates. Depending on what is being said an individual's words per minute will vary widely. But you could use a combination of a couple of PC programs to achieve your goal. First you'll need a good solid speech to text program. Feed it's results into a typing program that can track words per minute and your set.

Thank you mplichfamily - sorry for my slow reply, I have only just seen this. I will look into using separate programmes.
Thanks again

Yes, I'm looking at words per minute.

Come to think of it a good speech to text program may be able to tell you words per minute.