ReadAlign for Kotlin/fm.apakabar.readalign/SilenceHold

SilenceHold

Holds a word open through the silence behind it, and widens it into the quiet on both sides.

A recogniser marks where a word stops being audible, not where the voice has finished with it: the release of a final consonant and the fall of a line land past the mark, and the burst of an opening consonant lands before it.

Properties

frameSeconds

val frameSeconds: Double

Functions

energyFrames

fun energyFrames(samples: FloatArray, sampleRate: Double): List<Double>

How loud the recording is over each short stretch of it.

held

fun held(spans: List<WordSpan>, samples: FloatArray, sampleRate: Double, limit: Double = Rules.shared.holdLimit): List<WordSpan>

speechLevel

fun speechLevel(frames: List<Double>): Double

How loud this recording speaks, as the mean of its loudest share.

speechThreshold

fun speechThreshold(frames: List<Double>): Double

Louder than the room, quieter than a voice, taken from this recording rather than fixed: one reader is recorded hotter than another.

widened

fun widened(spans: List<WordSpan>, samples: FloatArray, sampleRate: Double, limit: Double = Rules.shared.holdLimit): List<WordSpan>

Each word given the quiet on both sides of it, so that a cut along its edges keeps the consonants a recogniser leaves outside its mark.