feat(audio): microphone voice activation

adding volume based voice detection

src/audio/audio.cpp

@@ -161,7 +161,15 @@
  * @fn void Audio::setInputGain(qreal dB)
  * @brief set the input gain
  *
- * @param[in] dB the new input gain in dB
+ * @fn void Audio::getInputThreshold()
+ * @brief get the current input threshold
+ *
+ * @return current input threshold percentage
+ *
+ * @fn void Audio::setInputThreshold(qreal percent)
+ * @brief set the input threshold
+ *
+ * @param[in] percent the new input threshold percentage
  */
 
 /**

src/audio/audio.h

@@ -72,6 +72,15 @@ public:
     virtual qreal inputGain() const = 0;
     virtual void setInputGain(qreal dB) = 0;
 
+    virtual qreal minInputThreshold() const = 0;
+    virtual void setMinInputThreshold(qreal dB) = 0;
+
+    virtual qreal maxInputThreshold() const = 0;
+    virtual void setMaxInputThreshold(qreal dB) = 0;
+
+    virtual qreal getInputThreshold() const = 0;
+    virtual void setInputThreshold(qreal percent) = 0;
+
     virtual void reinitInput(const QString& inDevDesc) = 0;
     virtual bool reinitOutput(const QString& outDevDesc) = 0;
 
@@ -91,6 +100,8 @@ public:
     virtual void playMono16Sound(const QByteArray& data) = 0;
     virtual void playMono16Sound(const QString& path) = 0;
 
+    virtual void stopActive() = 0;
+
     virtual void playAudioBuffer(uint sourceId, const int16_t* data, int samples, unsigned channels,
                                  int sampleRate) = 0;
 
@@ -104,6 +115,8 @@ protected:
 signals:
     void frameAvailable(const int16_t* pcm, size_t sample_count, uint8_t channels,
                         uint32_t sampling_rate);
+    void volumeAvailable(float value);
+    void startActive(qreal msec);
 };
 
 #endif // AUDIO_H

src/audio/backend/openal.cpp

@@ -57,6 +57,9 @@ OpenAL::OpenAL()
     , outputInitialized{false}
     , minInGain{-30}
     , maxInGain{30}
+    , minInThreshold{0.0f}
+    , maxInThreshold{0.4f}
+    , isActive{false}
 {
     // initialize OpenAL error stack
     alGetError();
@@ -67,6 +70,10 @@ OpenAL::OpenAL()
 
     moveToThread(audioThread);
 
+    voiceTimer.setSingleShot(true);
+    connect(this, &Audio::startActive, &voiceTimer, static_cast<void (QTimer::*)(int)>(&QTimer::start));
+    connect(&voiceTimer, &QTimer::timeout, this, &Audio::stopActive);
+
     connect(&captureTimer, &QTimer::timeout, this, &OpenAL::doCapture);
     captureTimer.setInterval(AUDIO_FRAME_DURATION / 2);
     captureTimer.setSingleShot(false);
@@ -175,6 +182,50 @@ void OpenAL::setMaxInputGain(qreal dB)
     maxInGain = dB;
 }
 
+/**
+ * @brief The minimum threshold value for an input device.
+ *
+ * @return minimum threshold percentage
+ */
+qreal OpenAL::minInputThreshold() const
+{
+    QMutexLocker locker(&audioLock);
+    return minInThreshold;
+}
+
+/**
+ * @brief Set the minimum allowed threshold percentage
+ *
+ * @note Default is 0%; usually you don't need to alter this value;
+ */
+void OpenAL::setMinInputThreshold(qreal percent)
+{
+    QMutexLocker locker(&audioLock);
+    minInThreshold = percent;
+}
+
+/**
+ * @brief The maximum threshold value for an input device.
+ *
+ * @return maximum threshold percentage
+ */
+qreal OpenAL::maxInputThreshold() const
+{
+    QMutexLocker locker(&audioLock);
+    return maxInThreshold;
+}
+
+/**
+ * @brief Set the maximum allowed threshold percentage
+ *
+ * @note Default is 40%; usually you don't need to alter this value.
+ */
+void OpenAL::setMaxInputThreshold(qreal percent)
+{
+    QMutexLocker locker(&audioLock);
+    maxInThreshold = percent;
+}
+
 void OpenAL::reinitInput(const QString& inDevDesc)
 {
     QMutexLocker locker(&audioLock);
@@ -278,6 +329,7 @@ bool OpenAL::initInput(const QString& deviceName, uint32_t channels)
     }
 
     setInputGain(Settings::getInstance().getAudioInGainDecibel());
+    setInputThreshold(Settings::getInstance().getAudioThreshold());
 
     qDebug() << "Opened audio input" << deviceName;
     alcCaptureStart(alInDev);
@@ -482,6 +534,36 @@ void OpenAL::playMono16SoundCleanup()
     }
 }
 
+/**
+ * @brief Called by doCapture to calculate volume of the audio buffer
+ *
+ * @param[in] buf   the current audio buffer
+ *
+ * @return volume in percent of max volume
+ */
+float OpenAL::getVolume(int16_t *buf)
+{
+    quint32 samples = AUDIO_FRAME_SAMPLE_COUNT * AUDIO_CHANNELS;
+    float sum = 0.0;
+    for (quint32 i = 0; i < samples; i++) {
+        float sample = (float)buf[i] / (float)std::numeric_limits<int16_t>::max();
+        if (sample > 0) {
+            sum += sample;
+        } else {
+            sum -= sample;
+        }
+    }
+    return sum/samples;
+}
+
+/**
+ * @brief Called by voiceTimer's timeout to disable audio broadcasting
+ */
+void OpenAL::stopActive()
+{
+    isActive = false;
+}
+
 /**
  * @brief Called on the captureTimer events to capture audio
  */
@@ -500,6 +582,19 @@ void OpenAL::doCapture()
     int16_t buf[AUDIO_FRAME_SAMPLE_COUNT * AUDIO_CHANNELS];
     alcCaptureSamples(alInDev, buf, AUDIO_FRAME_SAMPLE_COUNT);
 
+    float volume = getVolume(buf);
+    if (volume >= inputThreshold)
+    {
+        isActive = true;
+        emit startActive(voiceHold);
+    }
+
+    emit Audio::volumeAvailable(volume);
+    if (!isActive)
+    {
+        return;
+    }
+
     for (quint32 i = 0; i < AUDIO_FRAME_SAMPLE_COUNT * AUDIO_CHANNELS; ++i) {
         // gain amplification with clipping to 16-bit boundaries
         int ampPCM =
@@ -628,6 +723,11 @@ qreal OpenAL::inputGain() const
     return gain;
 }
 
+qreal OpenAL::getInputThreshold() const
+{
+    return inputThreshold;
+}
+
 qreal OpenAL::inputGainFactor() const
 {
     return gainFactor;
@@ -638,3 +738,9 @@ void OpenAL::setInputGain(qreal dB)
     gain = qBound(minInGain, dB, maxInGain);
     gainFactor = qPow(10.0, (gain / 20.0));
 }
+
+void OpenAL::setInputThreshold(qreal percent)
+{
+    inputThreshold = percent;
+}
+

src/audio/backend/openal.h

@@ -60,6 +60,15 @@ public:
     qreal inputGain() const;
     void setInputGain(qreal dB);
 
+    qreal minInputThreshold() const;
+    void setMinInputThreshold(qreal percent);
+
+    qreal maxInputThreshold() const;
+    void setMaxInputThreshold(qreal percent);
+
+    qreal getInputThreshold() const;
+    void setInputThreshold(qreal percent);
+
     void reinitInput(const QString& inDevDesc);
     bool reinitOutput(const QString& outDevDesc);
 
@@ -78,6 +87,7 @@ public:
     void stopLoop();
     void playMono16Sound(const QByteArray& data);
     void playMono16Sound(const QString& path);
+    void stopActive();
 
     void playAudioBuffer(uint sourceId, const int16_t* data, int samples, unsigned channels,
                          int sampleRate);
@@ -99,6 +109,7 @@ private:
     virtual bool initInput(const QString& deviceName);
     virtual bool initOutput(const QString& outDevDescr);
     void playMono16SoundCleanup();
+    float getVolume(int16_t *buf);
     void doCapture();
 
 protected:
@@ -120,6 +131,12 @@ protected:
     qreal gainFactor;
     qreal minInGain = -30;
     qreal maxInGain = 30;
+    qreal inputThreshold;
+    qreal voiceHold = 250;
+    bool isActive = false;
+    QTimer voiceTimer;
+    qreal minInThreshold = 0.0;
+    qreal maxInThreshold = 0.4;
 };
 
 #endif // OPENAL_H

src/audio/iaudiosettings.h

@@ -22,6 +22,9 @@ public:
     virtual qreal getAudioInGainDecibel() const = 0;
     virtual void setAudioInGainDecibel(qreal dB) = 0;
 
+    virtual qreal getAudioThreshold() const = 0;
+    virtual void setAudioThreshold(qreal percent) = 0;
+
     virtual int getOutVolume() const = 0;
     virtual void setOutVolume(int volume) = 0;
 
@@ -41,6 +44,7 @@ public:
     DECLARE_SIGNAL(audioOutDevEnabledChanged, bool enabled);
 
     DECLARE_SIGNAL(audioInGainDecibelChanged, qreal dB);
+    DECLARE_SIGNAL(audioThresholdChanged, qreal dB);
     DECLARE_SIGNAL(outVolumeChanged, int volume);
     DECLARE_SIGNAL(audioBitrateChanged, int bitrate);
     DECLARE_SIGNAL(enableTestSoundChanged, bool newValue);

src/persistence/settings.cpp

@@ -264,6 +264,7 @@ void Settings::loadGlobal()
         outDev = s.value("outDev", "").toString();
         audioOutDevEnabled = s.value("audioOutDevEnabled", true).toBool();
         audioInGainDecibel = s.value("inGain", 0).toReal();
+        audioThreshold = s.value("audioThreshold", 0).toReal();
         outVolume = s.value("outVolume", 100).toInt();
         audioBitrate = s.value("audioBitrate", 64).toInt();
         enableBackend2 = false;
@@ -566,6 +567,7 @@ void Settings::saveGlobal()
         s.setValue("outDev", outDev);
         s.setValue("audioOutDevEnabled", audioOutDevEnabled);
         s.setValue("inGain", audioInGainDecibel);
+        s.setValue("audioThreshold", audioThreshold);
         s.setValue("outVolume", outVolume);
         s.setValue("audioBitrate", audioBitrate);
         s.setValue("enableBackend2", enableBackend2);
@@ -1820,6 +1822,22 @@ void Settings::setAudioInGainDecibel(qreal dB)
     }
 }
 
+qreal Settings::getAudioThreshold() const
+{
+    QMutexLocker locker{&bigLock};
+    return audioThreshold;
+}
+
+void Settings::setAudioThreshold(qreal percent)
+{
+    QMutexLocker locker{&bigLock};
+
+    if (percent < audioThreshold || percent > audioThreshold) {
+        audioThreshold = percent;
+        emit audioThresholdChanged(audioThreshold);
+    }
+}
+
 QString Settings::getVideoDev() const
 {
     QMutexLocker locker{&bigLock};

src/persistence/settings.h

@@ -100,6 +100,8 @@ class Settings : public QObject, public ICoreSettings, public IAudioSettings, pu
                    audioInDevEnabledChanged FINAL)
     Q_PROPERTY(qreal audioInGainDecibel READ getAudioInGainDecibel WRITE setAudioInGainDecibel
                    NOTIFY audioInGainDecibelChanged FINAL)
+    Q_PROPERTY(qreal audioThreshold READ getAudioThreshold WRITE setAudioThreshold
+                   NOTIFY audioThresholdChanged FINAL)
     Q_PROPERTY(QString outDev READ getOutDev WRITE setOutDev NOTIFY outDevChanged FINAL)
     Q_PROPERTY(bool audioOutDevEnabled READ getAudioOutDevEnabled WRITE setAudioOutDevEnabled NOTIFY
                    audioOutDevEnabledChanged FINAL)
@@ -349,6 +351,9 @@ public:
     qreal getAudioInGainDecibel() const override;
     void setAudioInGainDecibel(qreal dB) override;
 
+    qreal getAudioThreshold() const override;
+    void setAudioThreshold(qreal percent) override;
+
     int getOutVolume() const override;
     void setOutVolume(int volume) override;
 
@@ -368,6 +373,7 @@ public:
     SIGNAL_IMPL(Settings, audioOutDevEnabledChanged, bool enabled)
 
     SIGNAL_IMPL(Settings, audioInGainDecibelChanged, qreal dB)
+    SIGNAL_IMPL(Settings, audioThresholdChanged, qreal percent)
     SIGNAL_IMPL(Settings, outVolumeChanged, int volume)
     SIGNAL_IMPL(Settings, audioBitrateChanged, int bitrate)
     SIGNAL_IMPL(Settings, enableTestSoundChanged, bool newValue)
@@ -632,6 +638,7 @@ private:
     QString inDev;
     bool audioInDevEnabled;
     qreal audioInGainDecibel;
+    qreal audioThreshold;
     QString outDev;
     bool audioOutDevEnabled;
     int outVolume;

src/widget/form/settings/avform.cpp

@@ -84,6 +84,18 @@ AVForm::AVForm(Audio* audio, CoreAV* coreAV, CameraSource& camera,
     microphoneSlider->setTracking(false);
     microphoneSlider->installEventFilter(this);
 
+    audioThresholdSlider->setToolTip(tr("Use slider to set the activation volume for your"
+                                    " input device."));
+    audioThresholdSlider->setMinimum(audio->minInputThreshold() * 1000);
+    audioThresholdSlider->setMaximum(audio->maxInputThreshold() * 1000);
+    audioThresholdSlider->setValue(audioSettings->getAudioThreshold() * 1000);
+    audioThresholdSlider->setTracking(false);
+    audioThresholdSlider->installEventFilter(this);
+
+    connect(audio, &Audio::volumeAvailable, this, &AVForm::setVolume);
+    volumeDisplay->setMinimum(audio->minInputThreshold() * 1000);
+    volumeDisplay->setMaximum(audio->maxInputThreshold() * 1000);
+
     fillAudioQualityComboBox();
 
     eventsInit();
@@ -149,6 +161,11 @@ void AVForm::rescanDevices()
     getVideoDevices();
 }
 
+void AVForm::setVolume(float value)
+{
+    volumeDisplay->setValue(value * 1000);
+}
+
 void AVForm::on_cbEnableBackend2_stateChanged()
 {
     audioSettings->setEnableBackend2(cbEnableBackend2->isChecked());
@@ -554,6 +571,14 @@ void AVForm::on_microphoneSlider_valueChanged(int value)
     audio->setInputGain(dB);
 }
 
+void AVForm::on_audioThresholdSlider_valueChanged(int value)
+{
+    const qreal percent = value / 1000.0;
+
+    audioSettings->setAudioThreshold(percent);
+    Audio::getInstance().setInputThreshold(percent);
+}
+
 void AVForm::createVideoSurface()
 {
     if (camVideoSurface)

src/widget/form/settings/avform.h

@@ -71,6 +71,7 @@ private slots:
     void on_playbackSlider_valueChanged(int value);
     void on_cbEnableTestSound_stateChanged();
     void on_microphoneSlider_valueChanged(int value);
+    void on_audioThresholdSlider_valueChanged(int value);
     void on_audioQualityComboBox_currentIndexChanged(int index);
 
     // camera
@@ -78,6 +79,7 @@ private slots:
     void on_videoModescomboBox_currentIndexChanged(int index);
 
     void rescanDevices();
+    void setVolume(float value);
 
     void on_cbEnableBackend2_stateChanged();
 

src/widget/form/settings/avform.ui

@@ -101,6 +101,37 @@
             </property>
            </widget>
           </item>
+          <item row="4" column="0">
+           <widget class="QLabel" name="audioThresholdLabel">
+            <property name="text">
+             <string>Threshold</string>
+            </property>
+           </widget>
+          </item>
+          <item row="4" column="1" colspan="2">
+           <widget class="QSlider" name="audioThresholdSlider">
+            <property name="orientation">
+             <enum>Qt::Horizontal</enum>
+            </property>
+           </widget>
+          </item>
+          <item row="5" column="0">
+           <widget class="QLabel" name="volumeDisplayLabel">
+            <property name="text">
+             <string>Volume</string>
+            </property>
+           </widget>
+          </item>
+          <item row="5" column="1" colspan="2">
+           <widget class="QProgressBar" name="volumeDisplay">
+            <property name="orientation">
+             <enum>Qt::Horizontal</enum>
+            </property>
+            <property name="textVisible">
+             <bool>false</bool>
+            </property>
+           </widget>
+          </item>
           <item row="0" column="0">
            <widget class="QLabel" name="outDevLabel">
             <property name="text">
@@ -108,14 +139,14 @@
             </property>
            </widget>
           </item>
-          <item row="4" column="0">
+          <item row="6" column="0">
            <widget class="QLabel" name="audioQualityLabel">
             <property name="text">
              <string>Audio quality</string>
             </property>
            </widget>
           </item>
-          <item row="4" column="1" colspan="2">
+          <item row="6" column="1" colspan="2">
            <widget class="QComboBox" name="audioQualityComboBox">
             <property name="sizePolicy">
              <sizepolicy hsizetype="Expanding" vsizetype="Fixed">
@@ -128,7 +159,7 @@
             </property>
            </widget>
           </item>
-          <item row="5" column="0" colspan="3">
+          <item row="7" column="0" colspan="3">
            <widget class="QCheckBox" name="cbEnableBackend2">
             <property name="enabled">
              <bool>true</bool>