[Feature] Voice Activity Detection improvement

[TobiasNeher]

Feature Description

Feature Description

This feature records the attached microphone and improves the detections when a person speaks

Definition of Completion

Main functions

• Live-records the attached microphone via the linux audio system (alsa?)
• For each audio sample, detects if a person is speaking (silero-vad)
• Buffers the samples if needed for debouncing
• Outputs the samples if speaking is detected

Speech controller internal interfaces

These interfaces are only used by nodes inside the speech controller package.
=> They don't need to be coordinated with the rest of the system

Testability

• Node only sends out audio samples when a person is speaking next to the robot/microphone
  • This might need some form of debouncing (e.g. speaking has to be detected at least 75% of the last second for sending)
• Node stops sending when the speaker has stopped for some amount of time (e.g. 3s)
• The end of the audio Stream has to be clearly defined

Code Dependencies

None

Specification Dependencies

• L01: Sprache erkennen

Definition of Completion

No response

Testability

No response

Code Dependencies

No response

Specification Dependencies

No response