Mirrors/overte
3
0
Fork 0
mirror of https://github.com/overte-org/overte.git synced 2025-08-07 22:30:43 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge pull request #9872 from ZappoMan/fadeInWhenOpeningNoiseGate

Browse source 
code review changes
This commit is contained in:
Brad Hefta-Gaub 2017-03-14 09:26:14 -07:00 committed by GitHub
commit a8d18adf83
3 changed files with 68 additions and 62 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
libraries/audio-client/src/AudioClient.cpp
View file

@ -1028,7 +1028,7 @@ void AudioClient::handleAudioInput() {
// if we performed the noise gate we can get values from it instead of enumerating the samples again // if we performed the noise gate we can get values from it instead of enumerating the samples again
_lastInputLoudness = _inputGate.getLastLoudness(); _lastInputLoudness = _inputGate.getLastLoudness();
if (_inputGate.clippedInLastFrame()) { if (_inputGate.clippedInLastBlock()) {
_timeSinceLastClip = 0.0f; _timeSinceLastClip = 0.0f;
} }
@ -1049,10 +1049,9 @@ void AudioClient::handleAudioInput() {
emit inputReceived({ reinterpret_cast<char*>(networkAudioSamples), numNetworkBytes }); emit inputReceived({ reinterpret_cast<char*>(networkAudioSamples), numNetworkBytes });
if (_inputGate.openedInLastFrame()) { if (_inputGate.openedInLastBlock()) {
emit noiseGateOpened(); emit noiseGateOpened();
} } else if (_inputGate.closedInLastBlock()) {
if (_inputGate.closedInLastFrame()) {
emit noiseGateClosed(); emit noiseGateClosed();
} }
@ -1072,7 +1071,7 @@ void AudioClient::handleAudioInput() {
// the output from the input gate (eventually, this could be crossfaded) // the output from the input gate (eventually, this could be crossfaded)
// and allow the codec to properly encode down to silent/zero. If we still // and allow the codec to properly encode down to silent/zero. If we still
// have _lastInputLoudness of 0 in our NEXT frame, we will send a silent packet // have _lastInputLoudness of 0 in our NEXT frame, we will send a silent packet
if (_lastInputLoudness == 0 && !_inputGate.closedInLastFrame()) { if (_lastInputLoudness == 0 && !_inputGate.closedInLastBlock()) {
packetType = PacketType::SilentAudioFrame; packetType = PacketType::SilentAudioFrame;
_silentOutbound.increment(); _silentOutbound.increment();
} else { } else {

97
libraries/audio-client/src/AudioNoiseGate.cpp
View file

@ -19,16 +19,16 @@
const float AudioNoiseGate::CLIPPING_THRESHOLD = 0.90f; const float AudioNoiseGate::CLIPPING_THRESHOLD = 0.90f;
AudioNoiseGate::AudioNoiseGate() : AudioNoiseGate::AudioNoiseGate() :
_inputFrameCounter(0), _inputBlockCounter(0),
_lastLoudness(0.0f), _lastLoudness(0.0f),
_quietestFrame(std::numeric_limits<float>::max()), _quietestBlock(std::numeric_limits<float>::max()),
_loudestFrame(0.0f), _loudestBlock(0.0f),
_didClipInLastFrame(false), _didClipInLastBlock(false),
_dcOffset(0.0f), _dcOffset(0.0f),
_measuredFloor(0.0f), _measuredFloor(0.0f),
_sampleCounter(0), _sampleCounter(0),
_isOpen(false), _isOpen(false),
_framesToClose(0) _blocksToClose(0)
{ {
} }
@ -37,7 +37,7 @@ void AudioNoiseGate::removeDCOffset(int16_t* samples, int numSamples) {
// //
// DC Offset correction // DC Offset correction
// //
// Measure the DC offset over a trailing number of frames, and remove it from the input signal. // Measure the DC offset over a trailing number of blocks, and remove it from the input signal.
// This causes the noise background measurements and server muting to be more accurate. Many off-board // This causes the noise background measurements and server muting to be more accurate. Many off-board
// ADC's have a noticeable DC offset. // ADC's have a noticeable DC offset.
// //
@ -51,7 +51,7 @@ void AudioNoiseGate::removeDCOffset(int16_t* samples, int numSamples) {
// Update measured DC offset // Update measured DC offset
measuredDcOffset /= numSamples; measuredDcOffset /= numSamples;
if (_dcOffset == 0.0f) { if (_dcOffset == 0.0f) {
// On first frame, copy over measured offset // On first block, copy over measured offset
_dcOffset = measuredDcOffset; _dcOffset = measuredDcOffset;
} else { } else {
_dcOffset = DC_OFFSET_AVERAGING * _dcOffset + (1.0f - DC_OFFSET_AVERAGING) * measuredDcOffset; _dcOffset = DC_OFFSET_AVERAGING * _dcOffset + (1.0f - DC_OFFSET_AVERAGING) * measuredDcOffset;
@ -69,13 +69,13 @@ void AudioNoiseGate::gateSamples(int16_t* samples, int numSamples) {
// //
// NOISE_GATE_HEIGHT: How loud you have to speak relative to noise background to open the gate. // NOISE_GATE_HEIGHT: How loud you have to speak relative to noise background to open the gate.
// Make this value lower for more sensitivity and less rejection of noise. // Make this value lower for more sensitivity and less rejection of noise.
// NOISE_GATE_WIDTH: The number of samples in an audio frame for which the height must be exceeded // NOISE_GATE_WIDTH: The number of samples in an audio block for which the height must be exceeded
// to open the gate. // to open the gate.
// NOISE_GATE_CLOSE_FRAME_DELAY: Once the noise is below the gate height for the frame, how many frames // NOISE_GATE_CLOSE_BLOCK_DELAY: Once the noise is below the gate height for the block, how many blocks
// will we wait before closing the gate. // will we wait before closing the gate.
// NOISE_GATE_FRAMES_TO_AVERAGE: How many audio frames should we average together to compute noise floor. // NOISE_GATE_BLOCKS_TO_AVERAGE: How many audio blocks should we average together to compute noise floor.
// More means better rejection but also can reject continuous things like singing. // More means better rejection but also can reject continuous things like singing.
// NUMBER_OF_NOISE_SAMPLE_FRAMES: How often should we re-evaluate the noise floor? // NUMBER_OF_NOISE_SAMPLE_BLOCKS: How often should we re-evaluate the noise floor?
float loudness = 0; float loudness = 0;
int thisSample = 0; int thisSample = 0;
@ -83,16 +83,16 @@ void AudioNoiseGate::gateSamples(int16_t* samples, int numSamples) {
const float NOISE_GATE_HEIGHT = 7.0f; const float NOISE_GATE_HEIGHT = 7.0f;
const int NOISE_GATE_WIDTH = 5; const int NOISE_GATE_WIDTH = 5;
const int NOISE_GATE_CLOSE_FRAME_DELAY = 5; const int NOISE_GATE_CLOSE_BLOCK_DELAY = 5;
const int NOISE_GATE_FRAMES_TO_AVERAGE = 5; const int NOISE_GATE_BLOCKS_TO_AVERAGE = 5;
// Check clipping, and check if should open noise gate // Check clipping, and check if should open noise gate
_didClipInLastFrame = false; _didClipInLastBlock = false;
for (int i = 0; i < numSamples; i++) { for (int i = 0; i < numSamples; i++) {
thisSample = std::abs(samples[i]); thisSample = std::abs(samples[i]);
if (thisSample >= ((float) AudioConstants::MAX_SAMPLE_VALUE * CLIPPING_THRESHOLD)) { if (thisSample >= ((float) AudioConstants::MAX_SAMPLE_VALUE * CLIPPING_THRESHOLD)) {
_didClipInLastFrame = true; _didClipInLastBlock = true;
} }
loudness += thisSample; loudness += thisSample;
@ -104,61 +104,64 @@ void AudioNoiseGate::gateSamples(int16_t* samples, int numSamples) {
_lastLoudness = fabs(loudness / numSamples); _lastLoudness = fabs(loudness / numSamples);
if (_quietestFrame > _lastLoudness) { if (_quietestBlock > _lastLoudness) {
_quietestFrame = _lastLoudness; _quietestBlock = _lastLoudness;
} }
if (_loudestFrame < _lastLoudness) { if (_loudestBlock < _lastLoudness) {
_loudestFrame = _lastLoudness; _loudestBlock = _lastLoudness;
} }
const int FRAMES_FOR_NOISE_DETECTION = 400; const int FRAMES_FOR_NOISE_DETECTION = 400;
if (_inputFrameCounter++ > FRAMES_FOR_NOISE_DETECTION) { if (_inputBlockCounter++ > FRAMES_FOR_NOISE_DETECTION) {
_quietestFrame = std::numeric_limits<float>::max(); _quietestBlock = std::numeric_limits<float>::max();
_loudestFrame = 0.0f; _loudestBlock = 0.0f;
_inputFrameCounter = 0; _inputBlockCounter = 0;
} }
// If Noise Gate is enabled, check and turn the gate on and off // If Noise Gate is enabled, check and turn the gate on and off
float averageOfAllSampleFrames = 0.0f; float averageOfAllSampleBlocks = 0.0f;
_sampleFrames[_sampleCounter++] = _lastLoudness; _sampleBlocks[_sampleCounter++] = _lastLoudness;
if (_sampleCounter == NUMBER_OF_NOISE_SAMPLE_FRAMES) { if (_sampleCounter == NUMBER_OF_NOISE_SAMPLE_BLOCKS) {
float smallestSample = std::numeric_limits<float>::max(); float smallestSample = std::numeric_limits<float>::max();
for (int i = 0; i <= NUMBER_OF_NOISE_SAMPLE_FRAMES - NOISE_GATE_FRAMES_TO_AVERAGE; i += NOISE_GATE_FRAMES_TO_AVERAGE) { for (int i = 0; i <= NUMBER_OF_NOISE_SAMPLE_BLOCKS - NOISE_GATE_BLOCKS_TO_AVERAGE; i += NOISE_GATE_BLOCKS_TO_AVERAGE) {
float thisAverage = 0.0f; float thisAverage = 0.0f;
for (int j = i; j < i + NOISE_GATE_FRAMES_TO_AVERAGE; j++) { for (int j = i; j < i + NOISE_GATE_BLOCKS_TO_AVERAGE; j++) {
thisAverage += _sampleFrames[j]; thisAverage += _sampleBlocks[j];
averageOfAllSampleFrames += _sampleFrames[j]; averageOfAllSampleBlocks += _sampleBlocks[j];
} }
thisAverage /= NOISE_GATE_FRAMES_TO_AVERAGE; thisAverage /= NOISE_GATE_BLOCKS_TO_AVERAGE;
if (thisAverage < smallestSample) { if (thisAverage < smallestSample) {
smallestSample = thisAverage; smallestSample = thisAverage;
} }
} }
averageOfAllSampleFrames /= NUMBER_OF_NOISE_SAMPLE_FRAMES; averageOfAllSampleBlocks /= NUMBER_OF_NOISE_SAMPLE_BLOCKS;
_measuredFloor = smallestSample; _measuredFloor = smallestSample;
_sampleCounter = 0; _sampleCounter = 0;
} }
_closedInLastFrame = false; _closedInLastBlock = false;
_openedInLastFrame = false; _openedInLastBlock = false;
if (samplesOverNoiseGate > NOISE_GATE_WIDTH) { if (samplesOverNoiseGate > NOISE_GATE_WIDTH) {
_openedInLastFrame = !_isOpen; _openedInLastBlock = !_isOpen;
_isOpen = true; _isOpen = true;
_framesToClose = NOISE_GATE_CLOSE_FRAME_DELAY; _blocksToClose = NOISE_GATE_CLOSE_BLOCK_DELAY;
} else { } else {
if (--_framesToClose == 0) { if (--_blocksToClose == 0) {
_closedInLastFrame = _isOpen; _closedInLastBlock = _isOpen;
_isOpen = false; _isOpen = false;
} }
} }
if (!_isOpen) { if (!_isOpen) {
if (_closedInLastFrame) { // First block after being closed gets faded to silence, we fade across
// would be nice to do a little crossfade to silence // the entire block on fading out. All subsequent blocks are muted by being slammed
// to zeros
if (_closedInLastBlock) {
float fadeSlope = (1.0f / numSamples);
for (int i = 0; i < numSamples; i++) { for (int i = 0; i < numSamples; i++) {
float fadedSample = (1.0f - (float)i / (float)numSamples) * (float)samples[i]; float fadedSample = (1.0f - ((float)i * fadeSlope)) * (float)samples[i];
samples[i] = (int16_t)fadedSample; samples[i] = (int16_t)fadedSample;
} }
} else { } else {
@ -167,10 +170,14 @@ void AudioNoiseGate::gateSamples(int16_t* samples, int numSamples) {
_lastLoudness = 0; _lastLoudness = 0;
} }
if (_openedInLastFrame) { if (_openedInLastBlock) {
// would be nice to do a little crossfade from silence // would be nice to do a little crossfade from silence, but we only want to fade
for (int i = 0; i < numSamples; i++) { // across the first 1/10th of the block, because we don't want to miss early
float fadedSample = ((float)i / (float)numSamples) * (float)samples[i]; // transients.
int fadeSamples = numSamples / 10; // fade over 1/10th of the samples
float fadeSlope = (1.0f / fadeSamples);
for (int i = 0; i < fadeSamples; i++) {
float fadedSample = (float)i * fadeSlope * (float)samples[i];
samples[i] = (int16_t)fadedSample; samples[i] = (int16_t)fadedSample;
} }
} }

24
libraries/audio-client/src/AudioNoiseGate.h
View file

@ -14,7 +14,7 @@
#include <stdint.h> #include <stdint.h>
const int NUMBER_OF_NOISE_SAMPLE_FRAMES = 300; const int NUMBER_OF_NOISE_SAMPLE_BLOCKS = 300;
class AudioNoiseGate { class AudioNoiseGate {
public: public:
@ -23,9 +23,9 @@ public:
void gateSamples(int16_t* samples, int numSamples); void gateSamples(int16_t* samples, int numSamples);
void removeDCOffset(int16_t* samples, int numSamples); void removeDCOffset(int16_t* samples, int numSamples);
bool clippedInLastFrame() const { return _didClipInLastFrame; } bool clippedInLastBlock() const { return _didClipInLastBlock; }
bool closedInLastFrame() const { return _closedInLastFrame; } bool closedInLastBlock() const { return _closedInLastBlock; }
bool openedInLastFrame() const { return _openedInLastFrame; } bool openedInLastBlock() const { return _openedInLastBlock; }
bool isOpen() const { return _isOpen; } bool isOpen() const { return _isOpen; }
float getMeasuredFloor() const { return _measuredFloor; } float getMeasuredFloor() const { return _measuredFloor; }
float getLastLoudness() const { return _lastLoudness; } float getLastLoudness() const { return _lastLoudness; }
@ -33,19 +33,19 @@ public:
static const float CLIPPING_THRESHOLD; static const float CLIPPING_THRESHOLD;
private: private:
int _inputFrameCounter; int _inputBlockCounter;
float _lastLoudness; float _lastLoudness;
float _quietestFrame; float _quietestBlock;
float _loudestFrame; float _loudestBlock;
bool _didClipInLastFrame; bool _didClipInLastBlock;
float _dcOffset; float _dcOffset;
float _measuredFloor; float _measuredFloor;
float _sampleFrames[NUMBER_OF_NOISE_SAMPLE_FRAMES]; float _sampleBlocks[NUMBER_OF_NOISE_SAMPLE_BLOCKS];
int _sampleCounter; int _sampleCounter;
bool _isOpen; bool _isOpen;
bool _closedInLastFrame { false }; bool _closedInLastBlock { false };
bool _openedInLastFrame { false }; bool _openedInLastBlock { false };
int _framesToClose; int _blocksToClose;
}; };
#endif // hifi_AudioNoiseGate_h #endif // hifi_AudioNoiseGate_h