n/futokb
1
0
Fork 0
mirror of https://gitlab.futo.org/keyboard/latinime.git synced 2024-09-28 14:54:30 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

Switch to PCM16 recording

Browse Source
This commit is contained in:
Aleksandras Kostarevas 2024-05-25 20:56:43 +03:00
parent b7ce7842ed
commit 43c46da519
1 changed files with 6 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
voiceinput-shared/src/main/java/org/futo/voiceinput/shared/AudioRecognizer.kt
View File

@ -150,7 +150,7 @@ class AudioRecognizer(
MediaRecorder.AudioSource.VOICE_RECOGNITION, MediaRecorder.AudioSource.VOICE_RECOGNITION,
16000, 16000,
AudioFormat.CHANNEL_IN_MONO, AudioFormat.CHANNEL_IN_MONO,
AudioFormat.ENCODING_PCM_FLOAT, AudioFormat.ENCODING_PCM_16BIT,
16000 * 2 * 5 16000 * 2 * 5
) )
@ -186,7 +186,7 @@ class AudioRecognizer(
var numConsecutiveNonSpeech = 0 var numConsecutiveNonSpeech = 0
var numConsecutiveSpeech = 0 var numConsecutiveSpeech = 0
val samples = FloatArray(1600) val samples = ShortArray(1600)
while (isRecording) { while (isRecording) {
yield() yield()
@ -226,14 +226,14 @@ class AudioRecognizer(
val samplesToRead = min(min(remainingSamples, 480), vadSampleBuffer.remaining()) val samplesToRead = min(min(remainingSamples, 480), vadSampleBuffer.remaining())
for (i in 0 until samplesToRead) { for (i in 0 until samplesToRead) {
vadSampleBuffer.put( vadSampleBuffer.put(
(samples[offset] * 32768.0).toInt().toShort() samples[offset]
) )
offset += 1 offset += 1
remainingSamples -= 1 remainingSamples -= 1
} }
} }
floatSamples.put(samples.sliceArray(0 until nRead)) floatSamples.put(samples.sliceArray(0 until nRead).map { it.toFloat() / Short.MAX_VALUE.toFloat() }.toFloatArray())
// Don't set hasTalked if the start sound may still be playing, otherwise on some // Don't set hasTalked if the start sound may still be playing, otherwise on some
// devices the rms just explodes and `hasTalked` is always true // devices the rms just explodes and `hasTalked` is always true
@ -243,7 +243,7 @@ class AudioRecognizer(
numConsecutiveNonSpeech = 0 numConsecutiveNonSpeech = 0
} }
val rms = sqrt(samples.sumOf { (it * it).toDouble() } / samples.size).toFloat() val rms = sqrt(samples.sumOf { (it.toFloat() / Short.MAX_VALUE.toFloat()).pow(2).toDouble() } / samples.size).toFloat()
if (startSoundPassed && ((rms > 0.01) || (numConsecutiveSpeech > 8))) { if (startSoundPassed && ((rms > 0.01) || (numConsecutiveSpeech > 8))) {
hasTalked = true hasTalked = true
@ -290,7 +290,7 @@ class AudioRecognizer(
} }
break break
} }
floatSamples.put(samples.sliceArray(0 until nRead2)) floatSamples.put(samples.sliceArray(0 until nRead2).map { it.toFloat() / Short.MAX_VALUE.toFloat() }.toFloatArray())
} else { } else {
break break
} }