|| || || || || ||

---

Mpeg7DocAudio.h

Go to the documentation of this file.

#ifndef Impala_Core_VideoSet_Mpeg7DocAudio_h
#define Impala_Core_VideoSet_Mpeg7DocAudio_h

#include "Util/XmlDoc.h"
#include "Util/QuickSort.h"
#include "Persistency/XmlFileReader.h"

namespace Impala
{
namespace Core
{
namespace VideoSet
{


class Mpeg7DocAudio : public Util::XmlDoc
{

public:

    Mpeg7DocAudio(String fileName, Util::IOBuffer* ioBuf)
    {
        ReadData(fileName, ioBuf);
    }

    virtual
    ~Mpeg7DocAudio()
    {
    }

    // inquiry

    int
    NrSegments()
    {
        return mSegmentId.size();
    }

    String
    SegmentId(int segment)
    {
        return mSegmentId[segment];
    }

    int
    StartFrame(int segment)
    {
        return mStartFrame[segment];
    }

    int
    StartFrameReFraction(int segment, int dstFraction)
    {
        return ReFraction(mStartFrame[segment], mFractions, dstFraction);
    }

    int
    EndFrame(int segment)
    {
        return mEndFrame[segment];
    }

    int
    EndFrameReFraction(int segment, int dstFraction)
    {
        return ReFraction(mEndFrame[segment], mFractions, dstFraction);
    }

    int
    NrAnno(int segment)
    {
        return mAnnos[segment].size();
    }

    String
    AnnoType(int segment, int anno)
    {
        return mAnnoTypes[segment][anno];
    }

    String
    Anno(int segment, int anno)
    {
        return mAnnos[segment][anno];
    }

    double
    Confidence(int segment)
    {
        return mConfidence[segment];
    }

    int
    FindSegment(int frameNr)
    {
        for (int i=0 ; i<mStartFrame.size() ; i++)
            if ((mStartFrame[i] <= frameNr) && (frameNr <= mEndFrame[i]))
                return i;
        return -1;
    }

    static int
    ReFraction(int srcFrame, int srcFraction, int dstFraction)
    {
        Int64 totalFractions = srcFrame;
        if (srcFraction == 30000) // probably NTSC with time unit PT1001N30000F
            totalFractions *= 1001;
        Int64 nrFractions = totalFractions % srcFraction;
        Int64 seconds = (totalFractions - nrFractions) / srcFraction;
        double f = (double) nrFractions / (double) srcFraction;
        nrFractions = f * dstFraction;
        return seconds * dstFraction + nrFractions;
    }

private:

    /* file excerpt:

<!-- Audio concept files look like this: -->

<?xml version="1.0" encoding="ISO-8859-1"?>
<Mpeg7 xmlns="urn:mpeg:mpeg7:schema:2001" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:mpeg7="urn:mpeg:mpeg7:schema:2001" xsi:schemaLocation="urn:mpeg:mpeg7:schema:2001 Mpeg7-2001.xsd">

<Description xsi:type="MediaDescriptionType">
  <MediaInformation>
    <MediaProfile>
      <MediaInstance>
        <InstanceIdentifier>Birds</InstanceIdentifier>
        <MediaLocator>
          <MediaUri>Birds.xml</MediaUri>
        </MediaLocator>
      </MediaInstance>
    </MediaProfile>
  </MediaInformation>
</Description>

<Description xsi:type="ContentEntityType">
  <MultimediaContent xsi:type="AudioType">
    <Audio>
      <MediaTime>
        <MediaTimePoint>T00:00:00:0F100</MediaTimePoint>
        <MediaDuration>PT0H40M54S7N100F</MediaDuration>
      </MediaTime>
      <TemporalDecomposition gap="true" overlap="false">

        <AudioSegment>
          <TextAnnotation confidence="0.746556">
            <FreeTextAnnotation>Birds</FreeTextAnnotation>
          </TextAnnotation>
          <MediaTime>
            <MediaTimePoint>T00:00:02:75F100</MediaTimePoint>
            <MediaDuration>PT0H0M39S75N100F</MediaDuration>
          </MediaTime>
        </AudioSegment>


<!-- Audio segmentation files look like this: -->

<?xml version="1.0" encoding="ISO-8859-1"?>
<Mpeg7 xmlns="urn:mpeg:mpeg7:schema:2001" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:mpeg7="urn:mpeg:mpeg7:schema:2001" xsi:schemaLocation="urn:mpeg:mpeg7:schema:2001 Mpeg7-2001.xsd">

<Description xsi:type="MediaDescriptionType">
  <MediaInformation>
    <MediaProfile>
      <MediaInstance>
        <InstanceIdentifier>
          Kosovosodo
        </InstanceIdentifier>
        <MediaLocator>
          <MediaUri>
            Kosovosodo.xml
          </MediaUri>
        </MediaLocator>
      </MediaInstance>
    </MediaProfile>
  </MediaInformation>
</Description>

<Description xsi:type="ContentEntityType">
  <MultimediaContent xsi:type="AudioType">
    <Audio>
      <MediaTime>
        <MediaTimePoint>T00:00:00:0F100</MediaTimePoint>
        <MediaDuration>PT0H40M54S7N100F</MediaDuration>
      </MediaTime>
      <TemporalDecomposition gap="true" overlap="false">

        <AudioSegment>
          <TextAnnotation type="background">
            <FreeTextAnnotation>Noise</FreeTextAnnotation>
          </TextAnnotation>
          <MediaTime>
            <MediaTimePoint>T00:00:00:0F100</MediaTimePoint>
            <MediaDuration>PT0H0M53S95N100F</MediaDuration>
          </MediaTime>
        </AudioSegment>
        <AudioSegment id="SPK1000-001">
          <TextAnnotation type="gender">
            <FreeTextAnnotation>Male</FreeTextAnnotation>
          </TextAnnotation>
          <TextAnnotation type="background">
            <FreeTextAnnotation>Noise</FreeTextAnnotation>
          </TextAnnotation>
          <MediaTime>
            <MediaTimePoint>T00:00:53:96F100</MediaTimePoint>
            <MediaDuration>PT0H0M2S95N100F</MediaDuration>
          </MediaTime>
        </AudioSegment>

    */

    void
    ReadData(String fileName, Util::IOBuffer* ioBuf)
    {
        Persistency::XmlFileReader reader;
        DOMDocument* doc = reader.Read(fileName, ioBuf);
        delete ioBuf;
        DOMNode* mpeg7 = GetChildNode(doc, "Mpeg7", true);
        DOMNode* desc1 = GetChildNode(mpeg7, "Description", true);
        DOMNode* desc2 = GetChildNodeAfter(mpeg7, desc1, "Description", true);
        DOMNode* mmContent = GetChildNode(desc2, "MultimediaContent", true);
        DOMNode* audio = GetChildNode(mmContent, "Audio", true);
        DOMNode* mediaTime = GetChildNode(audio, "MediaTime", true);
        mNrFrames = GetMediaDuration(mediaTime);
        ILOG_DEBUG("nrFrames = " << mNrFrames);
        mFractions = GetFractions(mediaTime);
        ILOG_DEBUG("fractions = " << mFractions);
        DOMNode* decomp = GetChildNode(audio, "TemporalDecomposition", true);
        std::vector<DOMNode*> segments = GetChildNodes(decomp, "AudioSegment");
        ILOG_DEBUG("nr AudioSegments = " << segments.size());
        for (int i=0 ; i<segments.size() ; i++)
            AddSegment(segments[i]);
    }

    void
    AddSegment(DOMNode* segment)
    {
        String id = GetAttributeValue(segment, "id", "<none>");
        ILOG_DEBUG("segment id = " << id);
        mSegmentId.push_back(id);
        std::vector<DOMNode*> annoNodes = GetChildNodes(segment,
                                                        "TextAnnotation");
        std::vector<String> annoTypes;
        std::vector<String> annos;
        for (int i=0 ; i<annoNodes.size() ; i++)
        {
            DOMNode* annoNode = annoNodes[i];
            String conf = GetAttributeValue(annoNode, "confidence", "1.0");
            mConfidence.push_back(atof(conf));
            DOMNode* freeText = GetChildNode(annoNode, "FreeTextAnnotation",
                                             true);
            String anno = GetElementValue(freeText);
            annos.push_back(anno);
            ILOG_DEBUG("anno = " << anno << " conf = " << conf);
        }
        mAnnoTypes.push_back(annoTypes);
        mAnnos.push_back(annos);

        DOMNode* mediaTime = GetChildNode(segment, "MediaTime", true);
        long frame = GetMediaTimePoint(mediaTime);
        long duration = GetMediaDuration(mediaTime);
        mStartFrame.push_back(frame);
        mEndFrame.push_back(frame + duration - 1);
    }

    long
    GetMediaTimePoint(DOMNode* mediaTime)
    {
        DOMNode* timePoint = GetChildNode(mediaTime, "MediaTimePoint", true);
        String s = GetElementValue(timePoint);
        // e.g. s = T00:28:21:5985F30000
        Util::StringParser p(s);
        p.Eat('T'); // T is begin of time point
        long hours = p.GetInt(':');
        long minutes = p.GetInt(':');
        long seconds = p.GetInt(':');
        p.AdvanceP(1); // skip ':', since it is not whitespace for next
        long fractions = p.GetInt('F');
        p.AdvanceP(1);
        long nrFractions = p.GetInt();
        long frame;
        frame = (hours*3600 + minutes*60 + seconds)*nrFractions + fractions;
        if (nrFractions == 30000) // probably NTSC with time unit PT1001N30000F
            frame = frame / 1001;
        return frame;
    }

    long
    GetMediaDuration(DOMNode* mediaTime)
    {
        DOMNode* duration = GetChildNode(mediaTime, "MediaDuration", true);
        String s = GetElementValue(duration);
        // e.g. s = PT00H00M04S2122N30000F
        Util::StringParser p(s);
        p.Eat('T'); // T is begin of duration
        long total = 0;
        if (p.Contains("H"))
        {
            total += p.GetInt('H') * 3600;
            p.AdvanceP(1);
        }
        if (p.Contains("M"))
        {
            total += p.GetInt('M') * 60;
            p.AdvanceP(1);
        }
        if (p.Contains("S"))
        {
            total += p.GetInt('S');
            p.AdvanceP(1);
        }
        long fractions = p.GetInt('N');
        p.AdvanceP(1);
        long nrFractions = p.GetInt('F');
        long dur = total*nrFractions + fractions;
        if (nrFractions == 30000) // probably NTSC with time unit PT1001N30000F
            dur /= 1001;
        return dur;
    }

    int
    GetFractions(DOMNode* mediaTime)
    {
        DOMNode* timePoint = GetChildNode(mediaTime, "MediaTimePoint", true);
        String s = GetElementValue(timePoint);
        // e.g. s = T00:28:21:5985F30000
        Util::StringParser p(s);
        p.Eat('F');
        return p.GetInt();
    }

    // data

    int    mNrFrames;
    int    mFractions;

    std::vector<String>                mSegmentId;
    std::vector<int>                   mStartFrame;
    std::vector<int>                   mEndFrame;
    std::vector< std::vector<String> > mAnnoTypes;
    std::vector< std::vector<String> > mAnnos;
    std::vector<double>                mConfidence;
    ILOG_VAR_DEC;

};

ILOG_VAR_INIT(Mpeg7DocAudio, Impala.Core.VideoSet);

} // namespace VideoSet
} // namespace Core
} // namespace Impala

#endif

---