ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

Speech Recognition Packages for ROS Kinetic Kame

asked 2016-10-23 15:35:24 -0500

canoski gravatar image

Hi, folks !

I need a speech recognition package for Kinetic Kame. I have looked at some packages like pocketsphinx but they are not for Kinetic Kame. Do you have any advice about how speech recognition works on Kinetic Kame ?

Thank you !

edit retag flag offensive close merge delete


I would be interested in this as well...any more information?

cerebraldad gravatar image cerebraldad  ( 2017-02-08 19:39:55 -0500 )edit

3 Answers

Sort by ยป oldest newest most voted

answered 2017-04-13 11:53:58 -0500

AndyZe gravatar image

updated 2017-09-18 09:01:37 -0500

As I see it, you have 2 alternatives. There's a Python version:

And a gstreamer version:

The Python version is supposedly faster and definitely easier to define new commands for. It uses an out-of-the-box CMUdict dictionary. Unfortunately, afaik, it can't be re-trained to match your specific needs.

The gstreamer version is a bit harder to use but it allows re-training, for example, if you have a strong accent.

There's also Google speech-to-text, which costs $ but its performance is amazing and it recognizes >100 languages/dialects.

edit flag offensive delete link more


Tried the version from mikeferguson, works for me.

Humpelstilzchen gravatar image Humpelstilzchen  ( 2017-04-27 04:32:53 -0500 )edit

That's good. I prefer my Python fork because it's easy to define your list of keywords and supposedly faster. Very good accuracy now that my sampling rate is correct, too. Should be a drop-in replacement for the original package.

AndyZe gravatar image AndyZe  ( 2017-04-27 07:54:15 -0500 )edit


I use Ros kinetix and ubuntu 16.0 but this link does not work for me. on the shell I just see some data but I can not recognize the words

I tried this link but does not work for me .

(INFO:n_live.c(105): Update to < 50.78 0.51 0

Developer gravatar image Developer  ( 2017-08-08 07:47:09 -0500 )edit

If you're seeing lines of output when you speak, it means Pocketsphinx is running but it may not recognize your speech. Do you have a strong accent? You can download different language models (e.g. Indian) here. link

AndyZe gravatar image AndyZe  ( 2017-08-08 10:53:18 -0500 )edit

thanks but I think it is not because of accent, because we have tried it with different accents. You re also right Pocketsphinx is running , but as I mentioned, I just see this interesting lines

INFO:n_live.c(105): Update form ..... INFO:n_live.c(105): Update to ....

Do I miss sthg. else ?

Developer gravatar image Developer  ( 2017-08-08 16:15:02 -0500 )edit

answered 2017-10-21 23:51:09 -0500

Shengye Wang gravatar image

If cloud is an option, you can try

edit flag offensive delete link more

answered 2017-10-07 12:27:40 -0500

PankajBaranwal gravatar image

Hey! Pocketsphinx is the right package for you.
But the original package is not well maintained and many have faced difficulties running it on Ubuntu14.04 and above, specially for ROS versions above Indigo.
So, you could instead use this one:
Updated Pocketsphinx It is well maintained, doesn't use GStreamer-like external dependencies, and has much better documentation.
UTNuclearRobotic's version is also a fork of a previous copy of this repo.
It has many modes you can work with including:
kws mode for keyword detection.
asr mode for continuous speech detection
okay-google mode for switching between the above mentioned modes smoothly etc.

You can also find more tutorials about it on the wiki as well as the Medium blogs here:

edit flag offensive delete link more

Question Tools

1 follower


Asked: 2016-10-23 15:35:24 -0500

Seen: 6,608 times

Last updated: Oct 21 '17