Added gain control and normalization, along with peak display

src/uk/co/majenko/audiobookrecorder/AudiobookRecorder.java

@@ -75,12 +75,14 @@ public class AudiobookRecorder extends JFrame {
     JScrollBar sampleScroll;
 
     JSpinner postSentenceGap;
+    JSpinner gainPercent;
     JCheckBox locked;
     JCheckBox attention;
     JCheckBox ethereal;
 
     JButtonSpacePlay reprocessAudioFFT;
     JButtonSpacePlay reprocessAudioPeak;
+    JButtonSpacePlay normalizeAudio;
 
     Thread playingThread = null;
 
@@ -363,6 +365,7 @@ public class AudiobookRecorder extends JFrame {
                     sampleWaveform.setMarkers(selectedSentence.getStartOffset(), selectedSentence.getEndOffset());
                     sampleWaveform.setAltMarkers(selectedSentence.getStartCrossing(), selectedSentence.getEndCrossing());
                     postSentenceGap.setValue(selectedSentence.getPostGap());
+                    gainPercent.setValue((int)(selectedSentence.getGain() * 100d));
                 }
             }
         });
@@ -375,12 +378,23 @@ public class AudiobookRecorder extends JFrame {
                     sampleWaveform.setMarkers(selectedSentence.getStartOffset(), selectedSentence.getEndOffset());
                     sampleWaveform.setAltMarkers(selectedSentence.getStartCrossing(), selectedSentence.getEndCrossing());
                     postSentenceGap.setValue(selectedSentence.getPostGap());
+                    gainPercent.setValue((int)(selectedSentence.getGain() * 100d));
                 }
             }
         });
 
+        normalizeAudio = new JButtonSpacePlay(Icons.normalize, "Normalize audio", new ActionListener() {
+            public void actionPerformed(ActionEvent e) {
+                if (selectedSentence != null) {
+                    selectedSentence.normalize();
+                    sampleWaveform.setData(selectedSentence.getAudioData());
+                }
+            }
+        });
+    
+
         postSentenceGap = new JSpinner(new SteppedNumericSpinnerModel(0, 5000, 100, 0));
-        postSentenceGap.setPreferredSize(new Dimension(75, 20));
+        postSentenceGap.setPreferredSize(new Dimension(50, 20));
 
         postSentenceGap.addChangeListener(new ChangeListener() {
             public void stateChanged(ChangeEvent e) {
@@ -391,6 +405,18 @@ public class AudiobookRecorder extends JFrame {
             }
         });
 
+        gainPercent = new JSpinner(new SteppedNumericSpinnerModel(0, 500, 1, 100));
+        gainPercent.setPreferredSize(new Dimension(50, 20));
+
+        gainPercent.addChangeListener(new ChangeListener() {
+            public void stateChanged(ChangeEvent e) {
+                JSpinner ob = (JSpinner)e.getSource();
+                if (selectedSentence != null) {
+                    selectedSentence.setGain((Integer)ob.getValue() / 100d);
+                    sampleWaveform.setData(selectedSentence.getAudioData());
+                }
+            }
+        });
 
         JToolBar controlsTop = new JToolBar(JToolBar.HORIZONTAL);
         JToolBar controlsLeft = new JToolBar(JToolBar.VERTICAL);
@@ -402,6 +428,7 @@ public class AudiobookRecorder extends JFrame {
 
         controlsLeft.add(reprocessAudioFFT);
         controlsLeft.add(reprocessAudioPeak);
+        controlsLeft.add(normalizeAudio);
 
         locked = new JCheckBox("Phrase locked");
         locked.setFocusable(false);
@@ -469,6 +496,11 @@ public class AudiobookRecorder extends JFrame {
         controlsTop.add(new JLabel("Post gap:"));
         controlsTop.add(postSentenceGap);
         controlsTop.add(new JLabel("ms"));
+
+        controlsTop.add(new JLabel("Gain:"));
+        controlsTop.add(gainPercent);
+        controlsTop.add(new JLabel("%"));
+
         controlsTop.add(Box.createHorizontalGlue());
 
 
@@ -1013,7 +1045,19 @@ public class AudiobookRecorder extends JFrame {
                     }
                 });
 
+                JMenuObject normalizeAll = new JMenuObject("Normalize chapter", c, new ActionListener() {
+                    public void actionPerformed(ActionEvent e) {
+                        JMenuObject o = (JMenuObject)e.getSource();
+                        Chapter c = (Chapter)o.getObject();
+                        for (Enumeration s = c.children(); s.hasMoreElements();) {
+                            Sentence snt = (Sentence)s.nextElement();
+                            snt.normalize();
+                        }
+                    }
+                });
+
                 menu.add(convertAll);
+                menu.add(normalizeAll);
                 menu.addSeparator();
                 menu.add(moveUp);
                 menu.add(moveDown);
@@ -1392,6 +1436,7 @@ public class AudiobookRecorder extends JFrame {
                 prefs.setProperty(String.format("%s.sentence.%08d.locked", keybase, i), snt.isLocked() ? "true" : "false");
                 prefs.setProperty(String.format("%s.sentence.%08d.attention", keybase, i), snt.getAttentionFlag() ? "true" : "false");
                 prefs.setProperty(String.format("%s.sentence.%08d.ethereal", keybase, i), snt.getEthereal() ? "true" : "false");
+                prefs.setProperty(String.format("%s.sentence.%08d.gain", keybase, i), String.format("%.8f", snt.getGain()));
                 i++;
             }
         }
@@ -1499,6 +1544,7 @@ public class AudiobookRecorder extends JFrame {
                     s.updateCrossings();
                     sampleWaveform.setAltMarkers(s.getStartCrossing(), s.getEndCrossing());
                     postSentenceGap.setValue(s.getPostGap());
+                    gainPercent.setValue((int)(s.getGain() * 100d));
                     locked.setSelected(s.isLocked());
                     attention.setSelected(s.getAttentionFlag());
                     ethereal.setSelected(s.getEthereal());
@@ -1510,6 +1556,7 @@ public class AudiobookRecorder extends JFrame {
                     selectedSentence = null;
                     sampleWaveform.clearData();
                     postSentenceGap.setValue(0);
+                    gainPercent.setValue(100);
                     locked.setSelected(false);
                     attention.setSelected(false);
                     ethereal.setSelected(false);
@@ -1553,6 +1600,7 @@ public class AudiobookRecorder extends JFrame {
             s.setLocked(Utils.s2b(prefs.getProperty(String.format("chapter.audition.sentence.%08d.locked", i))));
             s.setAttentionFlag(Utils.s2b(prefs.getProperty(String.format("chapter.audition.sentence.%08d.attention", i))));
             s.setEthereal(Utils.s2b(prefs.getProperty(String.format("chapter.audition.sentence.%08d.ethereal", i))));
+            s.setGain(Utils.s2d(prefs.getProperty(String.format("chapter.audition.sentence.%08d.gain", i))));
             bookTreeModel.insertNodeInto(s, c, c.getChildCount());
         }
 
@@ -1573,6 +1621,7 @@ public class AudiobookRecorder extends JFrame {
             s.setLocked(Utils.s2b(prefs.getProperty(String.format("chapter.open.sentence.%08d.locked", i))));
             s.setAttentionFlag(Utils.s2b(prefs.getProperty(String.format("chapter.open.sentence.%08d.attention", i))));
             s.setEthereal(Utils.s2b(prefs.getProperty(String.format("chapter.open.sentence.%08d.ethereal", i))));
+            s.setGain(Utils.s2d(prefs.getProperty(String.format("chapter.open.sentence.%08d.gain", i))));
             bookTreeModel.insertNodeInto(s, c, c.getChildCount());
         }
 
@@ -1599,6 +1648,7 @@ public class AudiobookRecorder extends JFrame {
                 s.setLocked(Utils.s2b(prefs.getProperty(String.format("chapter.%04d.sentence.%08d.locked", cno, i))));
                 s.setAttentionFlag(Utils.s2b(prefs.getProperty(String.format("chapter.%04d.sentence.%08d.attention", cno, i))));
                 s.setEthereal(Utils.s2b(prefs.getProperty(String.format("chapter.%04d.sentence.%08d.ethereal", cno, i))));
+                s.setGain(Utils.s2d(prefs.getProperty(String.format("chapter.%04d.sentence.%08d.gain", cno, i))));
                 bookTreeModel.insertNodeInto(s, c, c.getChildCount());
             }
         }
@@ -1620,6 +1670,7 @@ public class AudiobookRecorder extends JFrame {
             s.setLocked(Utils.s2b(prefs.getProperty(String.format("chapter.close.sentence.%08d.locked", i))));
             s.setAttentionFlag(Utils.s2b(prefs.getProperty(String.format("chapter.close.sentence.%08d.attention", i))));
             s.setEthereal(Utils.s2b(prefs.getProperty(String.format("chapter.close.sentence.%08d.ethereal", i))));
+            s.setGain(Utils.s2d(prefs.getProperty(String.format("chapter.close.sentence.%08d.gain", i))));
             bookTreeModel.insertNodeInto(s, c, c.getChildCount());
         }
 
@@ -1688,13 +1739,13 @@ public class AudiobookRecorder extends JFrame {
     }
 
     public int getNoiseFloorDB() {
-	int nf = getNoiseFloor();
+        int nf = getNoiseFloor();
-	if (nf == 0) return 0;
+        if (nf == 0) return 0;
-	double r = nf / 32767d;
+        double r = nf / 32767d;
-	double l10 = Math.log10(r);
+        double l10 = Math.log10(r);
-	double db = 20d * l10;
+        double db = 20d * l10;
 
-	return (int)db;
+        return (int)db;
     }
 
     public void recordRoomNoise() {

src/uk/co/majenko/audiobookrecorder/Icons.java

@@ -32,4 +32,5 @@ public class Icons {
     static public final ImageIcon zoomOut     = new ImageIcon(Icons.class.getResource("icons/zoom-out.png"));
     static public final ImageIcon dollar      = new ImageIcon(Icons.class.getResource("icons/dollar.png"));
     static public final ImageIcon attention   = new ImageIcon(Icons.class.getResource("icons/attention.png"));
+    static public final ImageIcon normalize   = new ImageIcon(Icons.class.getResource("icons/normalize.png"));
 }

src/uk/co/majenko/audiobookrecorder/Sentence.java

@@ -47,6 +47,8 @@ public class Sentence extends DefaultMutableTreeNode implements Cacheable {
     boolean inSample;
     boolean attention = false;
 
+    double gain = 1.0d;
+
     String havenJobId = "";
 
     // 0: Not processed
@@ -422,6 +424,10 @@ public class Sentence extends DefaultMutableTreeNode implements Cacheable {
     }
 
     public int[] getAudioDataS16LE(AudioInputStream s, AudioFormat format) throws IOException {
+        return getAudioDataS16LE(s, format, true);
+    }
+
+    public int[] getAudioDataS16LE(AudioInputStream s, AudioFormat format, boolean amplify) throws IOException {
         long len = s.getFrameLength();
         int frameSize = format.getFrameSize();
         int chans = format.getChannels();
@@ -441,7 +447,12 @@ public class Sentence extends DefaultMutableTreeNode implements Cacheable {
             } else {
                 sample = (frame[1] << 8) | frame[0];
             }
-            samples[(int)fno] = sample;
+            if (amplify) {
+                double amped = (double)sample * gain;
+                samples[(int)fno] = (int)amped;
+            } else {
+                samples[(int)fno] = sample;
+            }
         }
 
         return samples;
@@ -475,6 +486,28 @@ public class Sentence extends DefaultMutableTreeNode implements Cacheable {
         return null;
     }
 
+    public int[] getUnprocessedAudioData() {
+        File f = getFile();
+        try {
+            AudioInputStream s = AudioSystem.getAudioInputStream(f);
+            AudioFormat format = getAudioFormat();
+
+            int[] samples = null;
+
+            switch (format.getSampleSizeInBits()) {
+                case 16:
+                    samples = getAudioDataS16LE(s, format, false);
+                    break;
+            }
+
+            s.close();
+            return samples;
+        } catch (Exception e) {
+        }
+        return null;
+    }
+
+
     public int getStartCrossing() {
         return crossStartOffset;
     }
@@ -606,67 +639,108 @@ public class Sentence extends DefaultMutableTreeNode implements Cacheable {
         return null;
     }
 
-    byte[] postProcessData(byte[] data) {
+    byte[] adjustGain(byte[] data) {
-        if (effectEthereal) {
+        AudioFormat format = getAudioFormat();
-            AudioFormat format = getAudioFormat();
+        int frameSize = format.getFrameSize();
-            int frameSize = format.getFrameSize();
+        int channels = format.getChannels();
-            int channels = format.getChannels();
+        int bytesPerChannel = frameSize / channels;
-            int bytesPerChannel = frameSize / channels;
 
-            int frames = data.length / frameSize;
+        int frames = data.length / frameSize;
 
-            int byteNo = 0;
+        int byteNo = 0;
 
-            double fpms = (double)format.getFrameRate() / 1000d;
+        byte[] out = new byte[data.length];
-            double doubleOffset = fpms * (double) AudiobookRecorder.window.book.getInteger("effects.ethereal.offset");
-            int offset = (int)doubleOffset;
-            double attenuation = 1d - ((double)AudiobookRecorder.window.book.getInteger("effects.ethereal.attenuation") / 100d);
 
-            int copies = AudiobookRecorder.window.book.getInteger("effects.ethereal.iterations");
+        for (int i = 0; i < frames; i++) {
+            if (channels == 1) {
+                int l = data[i * frameSize] >= 0 ? data[i * frameSize] : 256 + data[i * frameSize];
+                int h = data[(i * frameSize) + 1] >= 0 ? data[(i * frameSize) + 1] : 256 + data[(i * frameSize) + 1];
 
-            byte[] out = new byte[data.length];
+                int sample = (h << 8) | l;
+                if ((sample & 0x8000) == 0x8000) sample |= 0xFFFF0000;
 
-            for (int i = 0; i < frames; i++) {
+                double sampleDouble = (double)sample;
-                if (channels == 1) {
+                sampleDouble *= gain;
-                    int l = data[i * frameSize] >= 0 ? data[i * frameSize] : 256 + data[i * frameSize];
+                sample = (int)sampleDouble;
-                    int h = data[(i * frameSize) + 1] >= 0 ? data[(i * frameSize) + 1] : 256 + data[(i * frameSize) + 1];
-                    
-                    int sample = (h << 8) | l;
-                    if ((sample & 0x8000) == 0x8000) sample |= 0xFFFF0000;
 
-                    double sampleDouble = (double)sample;
+                if (sample > 32767) sample = 32767;
+                if (sample < -32768) sample = -32768;
+                out[i * frameSize] = (byte)(sample & 0xFF);
+                out[(i * frameSize) + 1] = (byte)((sample & 0xFF00) >> 8);
 
-                    int used = 0;
+            } else {
-                    for (int j = 0; j < copies; j++) {
+                return data;
-                        if (i + (j * offset) < frames) {
-                            used++;
-                            int lx = data[(i + (j * offset)) * frameSize] >= 0 ? data[(i + (j * offset)) * frameSize] : 256 + data[(i + (j * offset)) * frameSize];
-                            int hx = data[((i + (j * offset)) * frameSize) + 1] >= 0 ? data[((i + (j * offset)) * frameSize) + 1] : 256 + data[((i + (j * offset)) * frameSize) + 1];
-                            int futureSample = (hx << 8) | lx;
-                            if ((futureSample & 0x8000) == 0x8000) futureSample |= 0xFFFF0000;
-                            double futureDouble = (double)futureSample;
-                            for (int k = 0; k < copies; k++) {
-                                futureDouble *= attenuation;
-                            }
-                            sampleDouble  = mix(sampleDouble, futureDouble);
-                        }
-                    }
-                    sample = (int)sampleDouble;
-                    if (sample > 32767) sample = 32767;
-                    if (sample < -32768) sample = -32768;
-                    out[i * frameSize] = (byte)(sample & 0xFF); 
-                    out[(i * frameSize) + 1] = (byte)((sample & 0xFF00) >> 8);
-    
-                } else {
-                    return data;
-                }
             }
-
-            return out;
-            
-        } else {
-            return data;
         }
+
+        return out;
+    }
+
+    byte[] postProcessData(byte[] data) {
+        data = adjustGain(data);
+
+        if (effectEthereal) {
+            data = processEtherealEffect(data);
+        }
+        return data;
+    }
+
+    byte[] processEtherealEffect(byte[] data) {
+        AudioFormat format = getAudioFormat();
+        int frameSize = format.getFrameSize();
+        int channels = format.getChannels();
+        int bytesPerChannel = frameSize / channels;
+
+        int frames = data.length / frameSize;
+
+        int byteNo = 0;
+
+        double fpms = (double)format.getFrameRate() / 1000d;
+        double doubleOffset = fpms * (double) AudiobookRecorder.window.book.getInteger("effects.ethereal.offset");
+        int offset = (int)doubleOffset;
+        double attenuation = 1d - ((double)AudiobookRecorder.window.book.getInteger("effects.ethereal.attenuation") / 100d);
+
+        int copies = AudiobookRecorder.window.book.getInteger("effects.ethereal.iterations");
+
+        byte[] out = new byte[data.length];
+
+        for (int i = 0; i < frames; i++) {
+            if (channels == 1) {
+                int l = data[i * frameSize] >= 0 ? data[i * frameSize] : 256 + data[i * frameSize];
+                int h = data[(i * frameSize) + 1] >= 0 ? data[(i * frameSize) + 1] : 256 + data[(i * frameSize) + 1];
+                
+                int sample = (h << 8) | l;
+                if ((sample & 0x8000) == 0x8000) sample |= 0xFFFF0000;
+
+                double sampleDouble = (double)sample;
+
+                int used = 0;
+                for (int j = 0; j < copies; j++) {
+                    if (i + (j * offset) < frames) {
+                        used++;
+                        int lx = data[(i + (j * offset)) * frameSize] >= 0 ? data[(i + (j * offset)) * frameSize] : 256 + data[(i + (j * offset)) * frameSize];
+                        int hx = data[((i + (j * offset)) * frameSize) + 1] >= 0 ? data[((i + (j * offset)) * frameSize) + 1] : 256 + data[((i + (j * offset)) * frameSize) + 1];
+                        int futureSample = (hx << 8) | lx;
+                        if ((futureSample & 0x8000) == 0x8000) futureSample |= 0xFFFF0000;
+                        double futureDouble = (double)futureSample;
+                        for (int k = 0; k < copies; k++) {
+                            futureDouble *= attenuation;
+                        }
+                        sampleDouble  = mix(sampleDouble, futureDouble);
+                    }
+                }
+                sample = (int)sampleDouble;
+                if (sample > 32767) sample = 32767;
+                if (sample < -32768) sample = -32768;
+                out[i * frameSize] = (byte)(sample & 0xFF); 
+                out[(i * frameSize) + 1] = (byte)((sample & 0xFF00) >> 8);
+
+            } else {
+                return data;
+            }
+        }
+
+        return out;
     }
 
     public void recognise() {
@@ -691,7 +765,7 @@ public class Sentence extends DefaultMutableTreeNode implements Cacheable {
 
     public void clearCache() {
         storedAudioData = null;
-        System.gc();
+//        System.gc();
     }
 
     public boolean lockedInCache() {
@@ -926,4 +1000,44 @@ public class Sentence extends DefaultMutableTreeNode implements Cacheable {
         return z;
     }
 
+   public int getPeakValue() {
+        int[] samples = getUnprocessedAudioData();
+        if (samples == null) {
+            return 0;
+        }
+        int ms = 0;
+        for (int i = 0; i < samples.length; i++) {
+            if (Math.abs(samples[i]) > ms) {
+                ms = Math.abs(samples[i]);
+            }
+        }
+        return ms;
+    }
+
+    public int getHeadroom() {
+        int nf = getPeakValue();
+        if (nf == 0) return 0;
+        double r = nf / 32767d;
+        double l10 = Math.log10(r);
+        double db = 20d * l10;
+
+        return (int)db;
+    }
+
+    public void setGain(double g) {
+        if (g <= 0.0001d) g = 1.0d;
+        if (g == gain) return;
+        gain = g;
+        clearCache();
+    }
+
+    public double getGain() {
+        return gain;
+    }
+
+    public void normalize() {
+        int max = getPeakValue();
+        setGain(23192d / max);
+    }
+
 }

src/uk/co/majenko/audiobookrecorder/Utils.java

@@ -47,6 +47,14 @@ public class Utils {
         return 0.0f;
     }
 
+    public static double s2d(String s) {
+        try {
+            return Double.parseDouble(s);
+        } catch (Exception e) {
+        }
+        return 0.0d;
+    }
+
     public static void browse(String url) {
         if (Desktop.isDesktopSupported()) {
             try {

src/uk/co/majenko/audiobookrecorder/Waveform.java

@@ -105,12 +105,26 @@ public class Waveform extends JPanel implements MouseListener, MouseMotionListen
                 if (hcnt > 0) have /= hcnt;
                 if (lcnt > 0) lave /= lcnt;
 
+                boolean clip = false;
+
+                if (lmax > 32000) { // -3dB == 23198?
+                    clip = true;
+                }
+
+                if (hmax > 32000) { // -3dB
+                    clip = true;
+                }
+
                 hmax /= scale;
                 lmax /= scale;
                 have /= scale;
                 lave /= scale;
 
-                g.setColor(new Color(0, 20, 200));
+                if (clip) {
+                    g.setColor(new Color(200, 20, 0));
+                } else {
+                    g.setColor(new Color(0, 20, 200));
+                }
                 g.drawLine(n, h/2 + lmax, n, h/2 - hmax);
                 g.setColor(new Color(0, 100, 255));
                 g.drawLine(n, h/2 + (int)lave, n, h/2 - (int)have);