Voice Recognition

voice_recognition.py

Overview

voice_recognition.py is a text stub for ASR: it converts integer “utterance IDs” into canned sentences so downstream components (order-verifier, task-manager, etc.) can be developed and tested without a real speech recogniser.

Interfaces (strongly-typed, stateless)

Direction / Type

Topic

Semantics

Required (sub)

/{robot}/mic_channel

Encoded mic “utterance” IDs (std_msgs/Int32)

Required (sub)

/mic_channel

Legacy mic channel (std_msgs/Int32)

Provided (pub)

/{robot}/voice_recogn

Recognised text (std_msgs/String)

Provided (pub)

/voice_recogn

Legacy speech topic, latched (std_msgs/String)

Contract

Pre-conditions

  • A voice_strings: dict[int, str] mapping is passed to the constructor.

Post-conditions

  • On each valid key, publishes exactly one sentence on both topics.

  • Unknown keys are ignored (no publish, last sentence remains latched).

Invariants

  • The node resets its internal key cache after a successful publish.

  • Publisher queues absorb short bursts (queue_size=10, latched for legacy).

Implementation summary

  1. Callbacks cache incoming Int32 messages to self.data.

  2. Main loop checks self.data at 10 Hz: a. If self.data matches a key in voice_strings, publishes the mapped

    sentence on both topics and logs it.

    1. Clears self.data to avoid replaying the same phrase.

  3. Holds no other state—perfect for hot-reload or unit tests.

class voice_recognition.VoiceRecognition(voice_strings: dict[int, str])[source]

Bases: object

Minimal dictionary-based ASR emulator.

Parameters:

voice_strings (dict[int, str]) – Maps Int32 .data values to sentences.

process_data() → None[source]

If cached self.data matches a key, publish the mapped sentence once.