Speech Processor

The Speech Processor effect in EasyEffects is powered by the SpeexDSP library, a lightweight and efficient audio processing toolkit originally derived from the Speex codec project. The Speech Processor provides real-time voice enhancement features suitable for telephony, voice chat, or low-resource environments.

State Settings

Noise Suppression

Reduces steady background noise such as fan hum or air conditioning. This feature uses spectral subtraction algorithms implemented by SpeexDSP to attenuate non-speech components while preserving intelligibility.

Automatic Gain Control (AGC)

Automatically adjusts the input gain to maintain consistent voice volume, regardless of the speaker’s distance or tone. This ensures a stable output level for speech clarity.

Voice Activity Detection (VAD)

Detects whether the incoming signal contains speech. When inactive (i.e., during silence or noise), processing is minimized to save resources and reduce unwanted output.

Dereverberation

Reduces reverberation caused by reflective environments (e.g., tiled rooms or large halls). This improves speech clarity by suppressing echo tails.

Voice Activity Probability Settings

Start

Sets the threshold for transitioning from non-speech to speech. A higher value makes detection more conservative.

Continue

Sets the threshold for remaining in the speech-detected state. A higher value reduces the chance of speech drop-out.

Noise Suppression Settings

Level

Defines how aggressively noise suppression is applied. Lower (more negative) values provide stronger noise removal but may affect voice quality.

Summary

The Speech Processor is a lightweight and versatile solution for real-time speech enhancement, especially suited for:

  • Voice calls

  • Online meetings

  • Embedded systems or low-power devices

Its combination of noise suppression, AGC, VAD, and dereverberation provides a practical and efficient toolchain for improving voice intelligibility in challenging environments.

References