gcloud_speech: Google Cloud Speech API ROS Integration

These packages integrate Google Cloud Speech API into ROS ecosystem.

Usage

gcloud_speech_msgs: ROS Message Definitions

gcloud_speech_msgs defines following messages:

• gcloud_speech_msgs/LinearPcm16Le16000Audio represents raw audio buffer in Linear PCM format, 16 bits, little endian, sample rate 16,000 Hz. It could be generated/converted by any other nodes.
• gcloud_speech_msgs/RecognitionHypothesis represents a recognition hypothesis. It is a pair of transcript and confidence score.
• gcloud_speech_msgs/SpeechToTextAction and other action server messages defines a actionlib interface to interact with Google Cloud Speech API.

gcloud_speech: Google Cloud Speech API ActionServer

gcloud_speech/gcloud_speech_action_node is a SimpleActionServer node. It could only serve one goal at a time (theoretically it could serve multiple goals, but we don’t have an immediate plan for this). Newer goal preempts older goal. A goal specifies hints, max listening time, max execution time, and some other parameters. And the action server will start streaming results from Google Cloud Speech API through action feedback. A goal can be preempted/canceled anytime, and the result will return a single transcript. We highly recommend relying on feedback to get recognition results.

gcloud_speech_action_node serves the action server at /cogrob/speech_to_text with type gcloud_speech_msgs/SpeechToTextAction. It listens to /cogrob/microphone_audio audio topic with type gcloud_speech_msgs/LinearPcm16Le16000Audio.

To start gcloud_speech_action_node, a few command line flags must be set correctly:

--gcloud_cred=/path/to/your/google_cloud_credential.json
--grpc_roots=/path/to/grpc_roots.pem

The grpc_roots.pem file can be downloaded from here. We have also included a copy in gcloud_speech_utils package under assets directory.

For Developers

We provide a speech_local_main program that does not rely on ROS. It uses portaudio (portaudio19-dev) to capture audio and prints recognition result to the screen. To build this program, enable set -DGCLOUD_SPEECH_LOCAL_MAIN=ON when invoking cmake (catkin build --cmake-args -DGCLOUD_SPEECH_LOCAL_MAIN=ON). Command line flag --mic=MicrophoneName sets the microphone the program uses.

gcloud_speech_utils: Utilities and Examples

This package provides a record_microphone_audio node that publishes LinearPcm16Le16000Audio. By default it looks for USB microphones. You can add --mic=MicrophoneName to override this option. The argument only need to partially match of the micrphone name seen by portaudio. The playback_microphone_audio.py script enables monitoring the microphone topic. Due to limitations of PortAudio and ROS, there is noticeable latency.

This package also provides a example_client.py program that interacts with the ActionServer.

You may also find some useful ROS launch files in this package.

Everything is gcloud_speech_utils package is experimental and for demostration purpose only. We recommend you NOT to use them in critical environment. Use these software at your own risk.

Google Cloud Console

Please visit https://console.cloud.google.com to add a project and create service accounts for the users under “IAM & admin”. We recommend one service account for each machine. You will be able to create keys (in JSON format) in the same console.

You must enable Speech API before you can use it. We have experienced no error message shown for Speech API not enabled error in the past. To enabled Speech API, search “Speech” in the search bar.

License

This package is licensed under the 3-clause BSD License.

Copyright ©(2017) The Regents of the University of California, All rights reserved.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

• Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
• Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
• Neither the name of the University of California nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.

File truncated at 100 lines see the full file