ua1zbe/meshcore-sar_android
1
0
Fork 0
mirror of https://github.com/dz0ny/meshcore-sar.git synced 2026-08-11 16:30:28 +00:00
Code Issues Packages Projects Releases 1 Wiki Activity

Fix voice quality regression

Browse Source
This commit is contained in:
Janez T
2026-03-12 19:38:41 +01:00
parent baa1aa6edd
commit d2e6b692f3
14 changed files with 1023 additions and 301 deletions
Download Patch File Download Diff File Expand all files Collapse all files

90
lib/screens/messages_tab.dart
View File

@@ -72,7 +72,7 @@ class _MessagesTabState extends State<MessagesTab> {
final VoiceRecorderService _voiceRecorder = VoiceRecorderService();
bool _isRecording = false;
bool _isSendingVoice = false;
static const int _maxVoicePackets = 10;
static const Duration _maxVoiceRecordingDuration = Duration(seconds: 4);
static const double _silenceRmsThreshold = 500.0;
static const double _silencePeakThreshold = 1400.0;
static const int _maxInteriorSilentChunks = 2;
@@ -865,9 +865,10 @@ class _MessagesTabState extends State<MessagesTab> {
final packetDuration = Duration(
milliseconds: _activeVoiceMode!.packetDurationMs,
);
final maxVoicePackets = _maxVoicePacketsForMode(_activeVoiceMode!);
debugPrint(
'🎙️ [Voice] session=$_currentVoiceSessionId mode=$_activeVoiceMode chunkDuration=${packetDuration.inMilliseconds}ms',
'🎙️ [Voice] session=$_currentVoiceSessionId mode=$_activeVoiceMode chunkDuration=${packetDuration.inMilliseconds}ms maxPackets=$maxVoicePackets',
);
_recordedChunks.clear();
@@ -877,9 +878,13 @@ class _MessagesTabState extends State<MessagesTab> {
final stream = _voiceRecorder.startCapture(
chunkDuration: packetDuration,
sampleRateHz: _activeVoiceMode!.sampleRateHz,
codecKind: _activeVoiceMode!.codec,
enableBandPassFilter: appProvider.isVoiceBandPassFilterEnabled,
enableCompressor: appProvider.isVoiceCompressorEnabled,
enableLimiter: appProvider.isVoiceLimiterEnabled,
enableAutoGain: appProvider.isVoiceAutoGainEnabled,
enableEchoCancellation: appProvider.isVoiceEchoCancellationEnabled,
enableNoiseSuppression: appProvider.isVoiceNoiseSuppressionEnabled,
);
debugPrint('🎙️ [Voice] capture started, listening for chunks...');
_voiceStreamSub = stream.listen(
@@ -890,7 +895,7 @@ class _MessagesTabState extends State<MessagesTab> {
'🎙️ [Voice] chunk #${_recordedChunks.length} received: ${pcmChunk.length} samples',
);
setState(() {});
if (_recordedChunks.length >= _maxVoicePackets) {
if (_recordedChunks.length >= maxVoicePackets) {
debugPrint('🎙️ [Voice] max packets reached, stopping');
_stopAndSendVoice();
}
@@ -907,6 +912,12 @@ class _MessagesTabState extends State<MessagesTab> {
}
}
int _maxVoicePacketsForMode(VoicePacketMode mode) {
final packets =
_maxVoiceRecordingDuration.inMilliseconds ~/ mode.packetDurationMs;
return packets < 1 ? 1 : packets;
}
Future<void> _stopAndSendVoice() async {
if (!_isRecording) return;
final trimSilenceEnabled = context
@@ -921,9 +932,14 @@ class _MessagesTabState extends State<MessagesTab> {
await _voiceRecorder.stopCapture();
final rawChunks = List<Int16List>.from(_recordedChunks);
final chunks = trimSilenceEnabled ? _trimSilence(rawChunks) : rawChunks;
final trimmedChunks = trimSilenceEnabled
? _trimSilence(rawChunks)
: rawChunks;
final sessionId = _currentVoiceSessionId;
final mode = _activeVoiceMode;
final chunks = mode == null
? trimmedChunks
: _prepareChunksForSending(trimmedChunks, mode);
_recordedChunks.clear();
debugPrint(
@@ -1005,10 +1021,15 @@ class _MessagesTabState extends State<MessagesTab> {
debugPrint(
'🎙️ [Voice] encoding $total packets for deferred voice fetch, mode=${mode.label}, session=$sessionId',
);
final encodedChunks = mode.codec == VoiceCodecKind.lpcnet
? await _encodeLpcNetChunks(codec, chunks, mode)
: <Uint8List>[];
for (var i = 0; i < total; i++) {
if (!mounted) return;
try {
final codec2Data = await codec.encode(chunks[i], mode);
final codec2Data = mode.codec == VoiceCodecKind.lpcnet
? encodedChunks[i]
: await codec.encode(chunks[i], mode);
debugPrint(
'🎙️ [Voice] packet $i/$total encoded: ${codec2Data.length} bytes',
);
@@ -1112,6 +1133,43 @@ class _MessagesTabState extends State<MessagesTab> {
messagesProvider.markMessageSent(msgId, 0, 0);
}
Future<List<Uint8List>> _encodeLpcNetChunks(
VoiceCodecService codec,
List<Int16List> chunks,
VoicePacketMode mode,
) async {
final totalSamples = chunks.fold<int>(
0,
(sum, chunk) => sum + chunk.length,
);
final merged = Int16List(totalSamples);
final encodedByteLengths = <int>[];
var sampleOffset = 0;
for (final chunk in chunks) {
merged.setRange(sampleOffset, sampleOffset + chunk.length, chunk);
sampleOffset += chunk.length;
encodedByteLengths.add((chunk.length ~/ 640) * 8);
}
final encoded = await codec.encode(merged, mode);
final encodedChunks = <Uint8List>[];
var byteOffset = 0;
for (final length in encodedByteLengths) {
encodedChunks.add(
Uint8List.sublistView(encoded, byteOffset, byteOffset + length),
);
byteOffset += length;
}
return encodedChunks;
}
List<Int16List> _prepareChunksForSending(
List<Int16List> chunks,
VoicePacketMode mode,
) {
return chunks;
}
List<Int16List> _trimSilence(List<Int16List> chunks) {
if (chunks.isEmpty) return chunks;
@@ -1141,16 +1199,28 @@ class _MessagesTabState extends State<MessagesTab> {
bool _isSilentChunk(Int16List chunk) {
if (chunk.isEmpty) return true;
final rms = _chunkRms(chunk);
final peak = _chunkPeak(chunk);
return rms < _silenceRmsThreshold && peak < _silencePeakThreshold;
}
double _chunkRms(Int16List chunk) {
var sumSquares = 0.0;
for (final sample in chunk) {
sumSquares += sample * sample;
}
return math.sqrt(sumSquares / chunk.length);
}
int _chunkPeak(Int16List chunk) {
var peak = 0;
for (final sample in chunk) {
final absSample = sample.abs();
if (absSample > peak) peak = absSample;
sumSquares += sample * sample;
if (absSample > peak) {
peak = absSample;
}
}
final rms = math.sqrt(sumSquares / chunk.length);
return rms < _silenceRmsThreshold && peak < _silencePeakThreshold;
return peak;
}
bool _isPublicChannelSelected() {