Mercurial > hg > audiodb
changeset 18:999c9c216565
minor revisions: mainly tidying up naming conventions (segment->track)
| author | mas01mc |
|---|---|
| date | Mon, 13 Aug 2007 19:14:33 +0000 |
| parents | 6d899df0cfe4 |
| children | 0c5884204732 |
| files | audioDB.cpp audioDB.h gengetopt.in |
| diffstat | 3 files changed, 319 insertions(+), 316 deletions(-) [+] |
line wrap: on
line diff
--- a/audioDB.cpp Fri Aug 10 04:52:33 2007 +0000 +++ b/audioDB.cpp Mon Aug 13 19:14:33 2007 +0000 @@ -50,9 +50,9 @@ -Q, --QUERY=searchtype content-based search on --database using --features as a query. Optionally restrict the - search to those segments identified in a + search to those tracks identified in a --keyList. (possible values="point", - "segment", "sequence") + "track", "sequence") -p, --qpoint=position ordinal position of query start point in --features file. (default=`0') -e, --exhaustive exhaustive search: iterate through all query @@ -61,7 +61,7 @@ -n, --pointnn=numpoints number of point nearest neighbours to use in retrieval. (default=`10') -R, --radius=DOUBLE radius search, returns all - points/segments/sequences inside given radius. + points/tracks/sequences inside given radius. (default=`1.0') -x, --expandfactor=DOUBLE time compress/expand factor of result length to query length [1.0 .. 100.0]. (default=`1.1') @@ -107,8 +107,8 @@ dbName(0), inFile(0), key(0), - segFile(0), - segFileName(0), + trackFile(0), + trackFileName(0), timesFile(0), timesFileName(0), usingTimes(0), @@ -121,8 +121,8 @@ queryType(O2_FLAG_POINT_QUERY), verbosity(1), pointNN(O2_DEFAULT_POINTNN), - segNN(O2_DEFAULT_SEGNN), - segTable(0), + trackNN(O2_DEFAULT_TRACKNN), + trackTable(0), fileTable(0), dataBuf(0), l2normTable(0), @@ -304,9 +304,9 @@ /* TO DO: REPLACE WITH if(args_info.keyList_given){ - segFileName=args_info.keyList_arg; - if(strlen(segFileName)>0 && !(segFile = new ifstream(segFileName,ios::in))) - error("Could not open keyList file for reading",segFileName); + trackFileName=args_info.keyList_arg; + if(strlen(trackFileName)>0 && !(trackFile = new ifstream(trackFileName,ios::in))) + error("Could not open keyList file for reading",trackFileName); } AND UPDATE BATCHINSERT() */ @@ -329,9 +329,9 @@ inFile=args_info.features_arg; if(args_info.keyList_given){ - segFileName=args_info.keyList_arg; - if(strlen(segFileName)>0 && !(segFile = new ifstream(segFileName,ios::in))) - error("Could not open keyList file for reading",segFileName); + trackFileName=args_info.keyList_arg; + if(strlen(trackFileName)>0 && !(trackFile = new ifstream(trackFileName,ios::in))) + error("Could not open keyList file for reading",trackFileName); } if(args_info.times_given){ @@ -344,8 +344,8 @@ } // query type - if(strncmp(args_info.QUERY_arg, "segment", MAXSTR)==0) - queryType=O2_FLAG_SEG_QUERY; + if(strncmp(args_info.QUERY_arg, "track", MAXSTR)==0) + queryType=O2_FLAG_TRACK_QUERY; else if(strncmp(args_info.QUERY_arg, "point", MAXSTR)==0) queryType=O2_FLAG_POINT_QUERY; else if(strncmp(args_info.QUERY_arg, "sequence", MAXSTR)==0) @@ -367,8 +367,8 @@ - segNN=args_info.resultlength_arg; - if(segNN<1 || segNN >10000) + trackNN=args_info.resultlength_arg; + if(trackNN<1 || trackNN >10000) error("resultlength out of range: 1 <= resultlength <= 1000"); @@ -395,13 +395,13 @@ --------------------------------------------------------------------------------- - keyTable : list of keys of segments + keyTable : list of keys of tracks -------------------------------------------------------------------------- | key 256 bytes | -------------------------------------------------------------------------- O2_MAXFILES*02_FILENAMELENGTH - segTable : Maps implicit feature index to a feature vector matrix + trackTable : Maps implicit feature index to a feature vector matrix -------------------------------------------------------------------------- | numVectors (4 bytes) | -------------------------------------------------------------------------- @@ -490,8 +490,8 @@ error("error reading db header"); fileTableOffset = O2_HEADERSIZE; - segTableOffset = fileTableOffset + O2_FILETABLESIZE*O2_MAXFILES; - dataoffset = segTableOffset + O2_SEGTABLESIZE*O2_MAXFILES; + trackTableOffset = fileTableOffset + O2_FILETABLESIZE*O2_MAXFILES; + dataoffset = trackTableOffset + O2_TRACKTABLESIZE*O2_MAXFILES; l2normTableOffset = O2_DEFAULTDBSIZE - O2_MAXFILES*O2_MEANNUMVECTORS*sizeof(double); timesTableOffset = l2normTableOffset - O2_MAXFILES*O2_MEANNUMVECTORS*sizeof(double); @@ -524,7 +524,7 @@ // Make some handy tables with correct types fileTable= (char*)(db+fileTableOffset); - segTable = (unsigned*)(db+segTableOffset); + trackTable = (unsigned*)(db+trackTableOffset); dataBuf = (double*)(db+dataoffset); l2normTable = (double*)(db+l2normTableOffset); timesTable = (double*)(db+timesTableOffset); @@ -558,7 +558,7 @@ return; } - // Make a segment index table of features to file indexes + // Make a track index table of features to file indexes unsigned numVectors = (statbuf.st_size-sizeof(int))/(sizeof(double)*dbH->dim); if(!numVectors){ if(verbosity) @@ -589,9 +589,9 @@ // Copy the header back to the database memcpy (db, dbH, sizeof(dbTableHeaderT)); - // Update segment to file index map - //memcpy (db+segTableOffset+(dbH->numFiles-1)*sizeof(unsigned), &numVectors, sizeof(unsigned)); - memcpy (segTable+dbH->numFiles-1, &numVectors, sizeof(unsigned)); + // Update track to file index map + //memcpy (db+trackTableOffset+(dbH->numFiles-1)*sizeof(unsigned), &numVectors, sizeof(unsigned)); + memcpy (trackTable+dbH->numFiles-1, &numVectors, sizeof(unsigned)); // Update the feature database memcpy (db+dataoffset+insertoffset, indata+sizeof(int), statbuf.st_size-sizeof(int)); @@ -691,8 +691,8 @@ error("Must use timestamps with timestamped database","use --times"); fileTableOffset = O2_HEADERSIZE; - segTableOffset = fileTableOffset + O2_FILETABLESIZE*O2_MAXFILES; - dataoffset = segTableOffset + O2_SEGTABLESIZE*O2_MAXFILES; + trackTableOffset = fileTableOffset + O2_FILETABLESIZE*O2_MAXFILES; + dataoffset = trackTableOffset + O2_TRACKTABLESIZE*O2_MAXFILES; l2normTableOffset = O2_DEFAULTDBSIZE - O2_MAXFILES*O2_MEANNUMVECTORS*sizeof(double); timesTableOffset = l2normTableOffset - O2_MAXFILES*O2_MEANNUMVECTORS*sizeof(double); @@ -734,7 +734,7 @@ // Make some handy tables with correct types fileTable= (char*)(db+fileTableOffset); - segTable = (unsigned*)(db+segTableOffset); + trackTable = (unsigned*)(db+trackTableOffset); dataBuf = (double*)(db+dataoffset); l2normTable = (double*)(db+l2normTableOffset); timesTable = (double*)(db+timesTableOffset); @@ -776,7 +776,7 @@ } else{ - // Make a segment index table of features to file indexes + // Make a track index table of features to file indexes unsigned numVectors = (statbuf.st_size-sizeof(int))/(sizeof(double)*dbH->dim); if(!numVectors){ if(verbosity) @@ -810,9 +810,9 @@ // Copy the header back to the database memcpy (db, dbH, sizeof(dbTableHeaderT)); - // Update segment to file index map - //memcpy (db+segTableOffset+(dbH->numFiles-1)*sizeof(unsigned), &numVectors, sizeof(unsigned)); - memcpy (segTable+dbH->numFiles-1, &numVectors, sizeof(unsigned)); + // Update track to file index map + //memcpy (db+trackTableOffset+(dbH->numFiles-1)*sizeof(unsigned), &numVectors, sizeof(unsigned)); + memcpy (trackTable+dbH->numFiles-1, &numVectors, sizeof(unsigned)); // Update the feature database memcpy (db+dataoffset+insertoffset, indata+sizeof(int), statbuf.st_size-sizeof(int)); @@ -861,14 +861,14 @@ soap_done(&soap); } -void audioDB::ws_query(const char*dbName, const char *segKey, const char* hostport){ +void audioDB::ws_query(const char*dbName, const char *trackKey, const char* hostport){ struct soap soap; adb__queryResult adbQueryResult; soap_init(&soap); if(soap_call_adb__query(&soap,hostport,NULL, - (char*)dbName,(char*)segKey,(char*)segFileName,(char*)timesFileName, - queryType, queryPoint, pointNN, segNN, sequenceLength, adbQueryResult)==SOAP_OK){ + (char*)dbName,(char*)trackKey,(char*)trackFileName,(char*)timesFileName, + queryType, queryPoint, pointNN, trackNN, sequenceLength, adbQueryResult)==SOAP_OK){ //std::cerr << "result list length:" << adbQueryResult.__sizeRlist << std::endl; for(int i=0; i<adbQueryResult.__sizeRlist; i++) std::cout << adbQueryResult.Rlist[i] << " " << adbQueryResult.Dist[i] @@ -903,9 +903,9 @@ unsigned dudCount=0; unsigned nullCount=0; for(unsigned k=0; k<dbH->numFiles; k++){ - if(segTable[k]<sequenceLength){ + if(trackTable[k]<sequenceLength){ dudCount++; - if(!segTable[k]) + if(!trackTable[k]) nullCount++; } } @@ -918,8 +918,8 @@ initTables(dbName,0); for(unsigned k=0, j=0; k<dbH->numFiles; k++){ - cout << fileTable+k*O2_FILETABLESIZE << " " << segTable[k] << endl; - j+=segTable[k]; + cout << fileTable+k*O2_FILETABLESIZE << " " << trackTable[k] << endl; + j+=trackTable[k]; } status(dbName); @@ -945,12 +945,12 @@ break; case O2_FLAG_SEQUENCE_QUERY: if(radius==0) - segSequenceQuery(dbName, inFile, adbQueryResult); + trackSequenceQuery(dbName, inFile, adbQueryResult); else - segSequenceQueryEuc(dbName, inFile, adbQueryResult); + trackSequenceQueryEuc(dbName, inFile, adbQueryResult); break; - case O2_FLAG_SEG_QUERY: - segPointQuery(dbName, inFile, adbQueryResult); + case O2_FLAG_TRACK_QUERY: + trackPointQuery(dbName, inFile, adbQueryResult); break; default: error("unrecognized queryType in query()"); @@ -1092,12 +1092,12 @@ // Loop over nearest neighbours for(k=0; k < pointNN; k++){ // Scan for key - unsigned cumSeg=0; + unsigned cumTrack=0; for(l=0 ; l<dbH->numFiles; l++){ - cumSeg+=segTable[l]; - if(sIndexes[k]<cumSeg){ + cumTrack+=trackTable[l]; + if(sIndexes[k]<cumTrack){ cout << fileTable+l*O2_FILETABLESIZE << " " << distances[k] << " " << qIndexes[k] << " " - << sIndexes[k]+segTable[l]-cumSeg << endl; + << sIndexes[k]+trackTable[l]-cumTrack << endl; break; } } @@ -1117,15 +1117,15 @@ adbQueryResult->Rlist[k]=new char[O2_MAXFILESTR]; adbQueryResult->Dist[k]=distances[k]; adbQueryResult->Qpos[k]=qIndexes[k]; - unsigned cumSeg=0; + unsigned cumTrack=0; for(l=0 ; l<dbH->numFiles; l++){ - cumSeg+=segTable[l]; - if(sIndexes[k]<cumSeg){ + cumTrack+=trackTable[l]; + if(sIndexes[k]<cumTrack){ sprintf(adbQueryResult->Rlist[k], "%s", fileTable+l*O2_FILETABLESIZE); break; } } - adbQueryResult->Spos[k]=sIndexes[k]+segTable[l]-cumSeg; + adbQueryResult->Spos[k]=sIndexes[k]+trackTable[l]-cumTrack; } } @@ -1144,15 +1144,15 @@ } -// segPointQuery -// return the segNN closest segs to the query seg -// uses average of pointNN points per seg -void audioDB::segPointQuery(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult){ +// trackPointQuery +// return the trackNN closest tracks to the query track +// uses average of pointNN points per track +void audioDB::trackPointQuery(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult){ initTables(dbName, inFile); // For each input vector, find the closest pointNN matching output vectors and report unsigned numVectors = (statbuf.st_size-sizeof(int))/(sizeof(double)*dbH->dim); - unsigned numSegs = dbH->numFiles; + unsigned numTracks = dbH->numFiles; double* query = (double*)(indata+sizeof(int)); double* data = dataBuf; @@ -1169,20 +1169,20 @@ } assert(pointNN>0 && pointNN<=O2_MAXNN); - assert(segNN>0 && segNN<=O2_MAXNN); + assert(trackNN>0 && trackNN<=O2_MAXNN); // Make temporary dynamic memory for results - double segDistances[segNN]; - unsigned segIDs[segNN]; - unsigned segQIndexes[segNN]; - unsigned segSIndexes[segNN]; + double trackDistances[trackNN]; + unsigned trackIDs[trackNN]; + unsigned trackQIndexes[trackNN]; + unsigned trackSIndexes[trackNN]; double distances[pointNN]; unsigned qIndexes[pointNN]; unsigned sIndexes[pointNN]; unsigned j=numVectors; // number of query points - unsigned k,l,n, seg, segOffset=0, processedSegs=0; + unsigned k,l,n, track, trackOffset=0, processedTracks=0; double thisDist; for(k=0; k<pointNN; k++){ @@ -1191,11 +1191,11 @@ sIndexes[k]=~0; } - for(k=0; k<segNN; k++){ - segDistances[k]=0.0; - segQIndexes[k]=~0; - segSIndexes[k]=~0; - segIDs[k]=~0; + for(k=0; k<trackNN; k++){ + trackDistances[k]=0.0; + trackQIndexes[k]=~0; + trackSIndexes[k]=~0; + trackIDs[k]=~0; } double meanQdur = 0; @@ -1222,7 +1222,7 @@ meanDBdur = new double[dbH->numFiles]; for(k=0; k<dbH->numFiles; k++){ meanDBdur[k]=0.0; - for(j=0; j<segTable[k]-1 ; j++) + for(j=0; j<trackTable[k]-1 ; j++) meanDBdur[k]+=timesTable[j+1]-timesTable[j]; meanDBdur[k]/=j; } @@ -1238,32 +1238,32 @@ numVectors=queryPoint+1; } - // build segment offset table - unsigned *segOffsetTable = new unsigned[dbH->numFiles]; - unsigned cumSeg=0; - unsigned segIndexOffset; + // build track offset table + unsigned *trackOffsetTable = new unsigned[dbH->numFiles]; + unsigned cumTrack=0; + unsigned trackIndexOffset; for(k=0; k<dbH->numFiles;k++){ - segOffsetTable[k]=cumSeg; - cumSeg+=segTable[k]*dbH->dim; + trackOffsetTable[k]=cumTrack; + cumTrack+=trackTable[k]*dbH->dim; } char nextKey[MAXSTR]; gettimeofday(&tv1, NULL); - for(processedSegs=0, seg=0 ; processedSegs < dbH->numFiles ; seg++, processedSegs++){ - if(segFile){ - if(!segFile->eof()){ - segFile->getline(nextKey,MAXSTR); - seg=getKeyPos(nextKey); + for(processedTracks=0, track=0 ; processedTracks < dbH->numFiles ; track++, processedTracks++){ + if(trackFile){ + if(!trackFile->eof()){ + trackFile->getline(nextKey,MAXSTR); + track=getKeyPos(nextKey); } else break; } - segOffset=segOffsetTable[seg]; // numDoubles offset - segIndexOffset=segOffset/dbH->dim; // numVectors offset + trackOffset=trackOffsetTable[track]; // numDoubles offset + trackIndexOffset=trackOffset/dbH->dim; // numVectors offset if(verbosity>7) - cerr << seg << "." << segOffset/(dbH->dim) << "." << segTable[seg] << " | ";cerr.flush(); + cerr << track << "." << trackOffset/(dbH->dim) << "." << trackTable[track] << " | ";cerr.flush(); if(dbH->flags & O2_FLAG_L2NORM) usingQueryPoint?query=queryCopy+queryPoint*dbH->dim:query=queryCopy; @@ -1274,8 +1274,8 @@ else j=numVectors; while(j--){ - k=segTable[seg]; // number of vectors in seg - data=dataBuf+segOffset; // data for seg + k=trackTable[track]; // number of vectors in track + data=dataBuf+trackOffset; // data for track while(k--){ thisDist=0; l=dbH->dim; @@ -1284,7 +1284,7 @@ thisDist+=*q++**data++; if(!usingTimes || (usingTimes - && fabs(meanDBdur[seg]-meanQdur)<meanQdur*timesTol)){ + && fabs(meanDBdur[track]-meanQdur)<meanQdur*timesTol)){ n=pointNN; while(n--){ if(thisDist>=distances[n]){ @@ -1297,7 +1297,7 @@ } distances[n]=thisDist; qIndexes[n]=numVectors-j-1; - sIndexes[n]=segTable[seg]-k-1; + sIndexes[n]=trackTable[track]-k-1; break; } } @@ -1305,32 +1305,32 @@ break; } } - } // seg + } // track // Move query pointer to next query point query+=dbH->dim; } // query - // Take the average of this seg's distance - // Test the seg distances + // Take the average of this track's distance + // Test the track distances thisDist=0; n=pointNN; while(n--) thisDist+=distances[pointNN-n-1]; thisDist/=pointNN; - n=segNN; + n=trackNN; while(n--){ - if(thisDist>=segDistances[n]){ - if((n==0 || thisDist<=segDistances[n-1])){ + if(thisDist>=trackDistances[n]){ + if((n==0 || thisDist<=trackDistances[n-1])){ // Copy all values above up the queue for( l=pointNN-1 ; l > n ; l--){ - segDistances[l]=segDistances[l-1]; - segQIndexes[l]=segQIndexes[l-1]; - segSIndexes[l]=segSIndexes[l-1]; - segIDs[l]=segIDs[l-1]; + trackDistances[l]=trackDistances[l-1]; + trackQIndexes[l]=trackQIndexes[l-1]; + trackSIndexes[l]=trackSIndexes[l-1]; + trackIDs[l]=trackIDs[l-1]; } - segDistances[n]=thisDist; - segQIndexes[n]=qIndexes[0]; - segSIndexes[n]=sIndexes[0]; - segIDs[n]=seg; + trackDistances[n]=thisDist; + trackQIndexes[n]=qIndexes[0]; + trackSIndexes[n]=sIndexes[0]; + trackIDs[n]=track; break; } } @@ -1342,11 +1342,11 @@ qIndexes[k]=~0; sIndexes[k]=~0; } - } // segs + } // tracks gettimeofday(&tv2, NULL); if(verbosity>1) - cerr << endl << "processed segs :" << processedSegs + cerr << endl << "processed tracks :" << processedTracks << " elapsed time:" << ( tv2.tv_sec*1000 + tv2.tv_usec/1000 ) - ( tv1.tv_sec*1000+tv1.tv_usec/1000 ) << " msec" << endl; if(adbQueryResult==0){ @@ -1354,12 +1354,12 @@ cerr<<endl; // Output answer // Loop over nearest neighbours - for(k=0; k < min(segNN,processedSegs); k++) - cout << fileTable+segIDs[k]*O2_FILETABLESIZE - << " " << segDistances[k] << " " << segQIndexes[k] << " " << segSIndexes[k] << endl; + for(k=0; k < min(trackNN,processedTracks); k++) + cout << fileTable+trackIDs[k]*O2_FILETABLESIZE + << " " << trackDistances[k] << " " << trackQIndexes[k] << " " << trackSIndexes[k] << endl; } else{ // Process Web Services Query - int listLen = min(segNN, processedSegs); + int listLen = min(trackNN, processedTracks); adbQueryResult->__sizeRlist=listLen; adbQueryResult->__sizeDist=listLen; adbQueryResult->__sizeQpos=listLen; @@ -1370,17 +1370,17 @@ adbQueryResult->Spos = new int[listLen]; for(k=0; k<adbQueryResult->__sizeRlist; k++){ adbQueryResult->Rlist[k]=new char[O2_MAXFILESTR]; - adbQueryResult->Dist[k]=segDistances[k]; - adbQueryResult->Qpos[k]=segQIndexes[k]; - adbQueryResult->Spos[k]=segSIndexes[k]; - sprintf(adbQueryResult->Rlist[k], "%s", fileTable+segIDs[k]*O2_FILETABLESIZE); + adbQueryResult->Dist[k]=trackDistances[k]; + adbQueryResult->Qpos[k]=trackQIndexes[k]; + adbQueryResult->Spos[k]=trackSIndexes[k]; + sprintf(adbQueryResult->Rlist[k], "%s", fileTable+trackIDs[k]*O2_FILETABLESIZE); } } // Clean up - if(segOffsetTable) - delete segOffsetTable; + if(trackOffsetTable) + delete trackOffsetTable; if(queryCopy) delete queryCopy; if(qNorm) @@ -1396,17 +1396,17 @@ } -// NBest matched filter distance between query and target segs +// NBest matched filter distance between query and target tracks // efficient implementation // outputs average of N minimum matched filter distances -void audioDB::segSequenceQuery(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult){ +void audioDB::trackSequenceQuery(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult){ initTables(dbName, inFile); // For each input vector, find the closest pointNN matching output vectors and report // we use stdout in this stub version unsigned numVectors = (statbuf.st_size-sizeof(int))/(sizeof(double)*dbH->dim); - unsigned numSegs = dbH->numFiles; + unsigned numTracks = dbH->numFiles; double* query = (double*)(indata+sizeof(int)); double* data = dataBuf; @@ -1443,14 +1443,14 @@ memcpy(sNorm, l2normTable, dbVectors*sizeof(double)); double* snPtr = sNorm; for(i=0; i<dbH->numFiles; i++){ - if(segTable[i]>sequenceLength){ + if(trackTable[i]>sequenceLength){ tmp1=*snPtr; j=1; w=sequenceLength-1; while(w--) *snPtr+=snPtr[j++]; ps = snPtr+1; - w=segTable[i]-sequenceLength; // +1 - 1 + w=trackTable[i]-sequenceLength; // +1 - 1 while(w--){ tmp2=*ps; *ps=*(ps-1)-tmp1+*(ps+sequenceLength); @@ -1458,7 +1458,7 @@ ps++; } } - snPtr+=segTable[i]; + snPtr+=trackTable[i]; } double* pn = sMeanL2; @@ -1466,22 +1466,22 @@ while(w--) *pn++=0.0; ps=sNorm; - unsigned processedSegs=0; + unsigned processedTracks=0; for(i=0; i<dbH->numFiles; i++){ - if(segTable[i]>sequenceLength-1){ - w = segTable[i]-sequenceLength+1; + if(trackTable[i]>sequenceLength-1){ + w = trackTable[i]-sequenceLength+1; pn = sMeanL2+i; while(w--) *pn+=*ps++; - *pn/=segTable[i]-sequenceLength+1; + *pn/=trackTable[i]-sequenceLength+1; SILENCE_THRESH+=*pn; - processedSegs++; + processedTracks++; } - ps = sNorm + segTable[i]; + ps = sNorm + trackTable[i]; } if(verbosity>1) - cerr << "processedSegs: " << processedSegs << endl; - SILENCE_THRESH/=processedSegs; + cerr << "processedTracks: " << processedTracks << endl; + SILENCE_THRESH/=processedTracks; USE_THRESH=1; // Turn thresholding on DIFF_THRESH=SILENCE_THRESH/2; // 50% of the mean shingle power SILENCE_THRESH/=10; // 10% of the mean shingle power is SILENCE @@ -1507,23 +1507,23 @@ if(verbosity>1) - cerr << "matching segs..." << endl; + cerr << "matching tracks..." << endl; assert(pointNN>0 && pointNN<=O2_MAXNN); - assert(segNN>0 && segNN<=O2_MAXNN); + assert(trackNN>0 && trackNN<=O2_MAXNN); // Make temporary dynamic memory for results - double segDistances[segNN]; - unsigned segIDs[segNN]; - unsigned segQIndexes[segNN]; - unsigned segSIndexes[segNN]; + double trackDistances[trackNN]; + unsigned trackIDs[trackNN]; + unsigned trackQIndexes[trackNN]; + unsigned trackSIndexes[trackNN]; double distances[pointNN]; unsigned qIndexes[pointNN]; unsigned sIndexes[pointNN]; - unsigned k,l,m,n,seg,segOffset=0, HOP_SIZE=sequenceHop, wL=sequenceLength; + unsigned k,l,m,n,track,trackOffset=0, HOP_SIZE=sequenceHop, wL=sequenceLength; double thisDist; double oneOverWL=1.0/wL; @@ -1533,11 +1533,11 @@ sIndexes[k]=~0; } - for(k=0; k<segNN; k++){ - segDistances[k]=0.0; - segQIndexes[k]=~0; - segSIndexes[k]=~0; - segIDs[k]=~0; + for(k=0; k<trackNN; k++){ + trackDistances[k]=0.0; + trackQIndexes[k]=~0; + trackSIndexes[k]=~0; + trackIDs[k]=~0; } // Timestamp and durations processing @@ -1569,7 +1569,7 @@ assert(meanDBdur); for(k=0; k<dbH->numFiles; k++){ meanDBdur[k]=0.0; - for(j=0; j<segTable[k]-1 ; j++) + for(j=0; j<trackTable[k]-1 ; j++) meanDBdur[k]+=timesTable[j+1]-timesTable[j]; meanDBdur[k]/=j; } @@ -1595,62 +1595,62 @@ assert(DD); gettimeofday(&tv1, NULL); - processedSegs=0; - unsigned successfulSegs=0; + processedTracks=0; + unsigned successfulTracks=0; double* qp; double* sp; double* dp; double diffL2; - // build segment offset table - unsigned *segOffsetTable = new unsigned[dbH->numFiles]; - unsigned cumSeg=0; - unsigned segIndexOffset; + // build track offset table + unsigned *trackOffsetTable = new unsigned[dbH->numFiles]; + unsigned cumTrack=0; + unsigned trackIndexOffset; for(k=0; k<dbH->numFiles;k++){ - segOffsetTable[k]=cumSeg; - cumSeg+=segTable[k]*dbH->dim; + trackOffsetTable[k]=cumTrack; + cumTrack+=trackTable[k]*dbH->dim; } char nextKey [MAXSTR]; - for(processedSegs=0, seg=0 ; processedSegs < dbH->numFiles ; seg++, processedSegs++){ + for(processedTracks=0, track=0 ; processedTracks < dbH->numFiles ; track++, processedTracks++){ - // get segID from file if using a control file - if(segFile){ - if(!segFile->eof()){ - segFile->getline(nextKey,MAXSTR); - seg=getKeyPos(nextKey); + // get trackID from file if using a control file + if(trackFile){ + if(!trackFile->eof()){ + trackFile->getline(nextKey,MAXSTR); + track=getKeyPos(nextKey); } else break; } - segOffset=segOffsetTable[seg]; // numDoubles offset - segIndexOffset=segOffset/dbH->dim; // numVectors offset + trackOffset=trackOffsetTable[track]; // numDoubles offset + trackIndexOffset=trackOffset/dbH->dim; // numVectors offset - if(sequenceLength<segTable[seg]){ // test for short sequences + if(sequenceLength<trackTable[track]){ // test for short sequences if(verbosity>7) - cerr << seg << "." << segIndexOffset << "." << segTable[seg] << " | ";cerr.flush(); + cerr << track << "." << trackIndexOffset << "." << trackTable[track] << " | ";cerr.flush(); // Cross-correlation matrix for(j=0; j<numVectors;j++){ - D[j]=new double[segTable[seg]]; + D[j]=new double[trackTable[track]]; assert(D[j]); } // Matched filter matrix for(j=0; j<numVectors;j++){ - DD[j]=new double[segTable[seg]]; + DD[j]=new double[trackTable[track]]; assert(DD[j]); } // Cross Correlation for(j=0; j<numVectors; j++) - for(k=0; k<segTable[seg]; k++){ + for(k=0; k<trackTable[track]; k++){ qp=query+j*dbH->dim; - sp=dataBuf+segOffset+k*dbH->dim; + sp=dataBuf+trackOffset+k*dbH->dim; DD[j][k]=0.0; // Initialize matched filter array dp=&D[j][k]; // point to correlation cell j,k *dp=0.0; // initialize correlation cell @@ -1667,7 +1667,7 @@ for(j=0; j<numVectors-w; j++){ sp=DD[j]; spd=D[j+w]+w; - k=segTable[seg]-w; + k=trackTable[track]-w; while(k--) *sp+++=*spd++; } @@ -1677,7 +1677,7 @@ for(j=0; j<numVectors-w; j+=HOP_SIZE){ sp=DD[j]; spd=D[j+w]+w; - for(k=0; k<segTable[seg]-w; k+=HOP_SIZE){ + for(k=0; k<trackTable[track]-w; k+=HOP_SIZE){ *sp+=*spd; sp+=HOP_SIZE; spd+=HOP_SIZE; @@ -1686,13 +1686,13 @@ } if(verbosity>3 && usingTimes){ - cerr << "meanQdur=" << meanQdur << " meanDBdur=" << meanDBdur[seg] << endl; + cerr << "meanQdur=" << meanQdur << " meanDBdur=" << meanDBdur[track] << endl; cerr.flush(); } if(!usingTimes || (usingTimes - && fabs(meanDBdur[seg]-meanQdur)<meanQdur*timesTol)){ + && fabs(meanDBdur[track]-meanQdur)<meanQdur*timesTol)){ if(verbosity>3 && usingTimes){ cerr << "within duration tolerance." << endl; @@ -1701,7 +1701,7 @@ // Search for minimum distance by shingles (concatenated vectors) for(j=0;j<numVectors-wL+1;j+=HOP_SIZE) - for(k=0;k<segTable[seg]-wL+1;k+=HOP_SIZE){ + for(k=0;k<trackTable[track]-wL+1;k+=HOP_SIZE){ diffL2 = fabs(qNorm[j] - sNorm[k]); // Power test @@ -1709,7 +1709,7 @@ // Threshold on mean L2 of Q and S sequences (USE_THRESH && qNorm[j]>SILENCE_THRESH && sNorm[k]>SILENCE_THRESH && // Are both query and target windows above mean energy? - (qNorm[j]>qMeanL2 && sNorm[k]>sMeanL2[seg] && diffL2 < DIFF_THRESH ))) + (qNorm[j]>qMeanL2 && sNorm[k]>sMeanL2[track] && diffL2 < DIFF_THRESH ))) thisDist=DD[j][k]*oneOverWL; else thisDist=0.0; @@ -1746,25 +1746,25 @@ // Let's see the distances then... if(verbosity>3) - cerr << fileTable+seg*O2_FILETABLESIZE << " " << thisDist << endl; + cerr << fileTable+track*O2_FILETABLESIZE << " " << thisDist << endl; - // All the seg stuff goes here - n=segNN; + // All the track stuff goes here + n=trackNN; while(n--){ - if(thisDist>=segDistances[n]){ - if((n==0 || thisDist<=segDistances[n-1])){ + if(thisDist>=trackDistances[n]){ + if((n==0 || thisDist<=trackDistances[n-1])){ // Copy all values above up the queue - for( l=segNN-1 ; l > n ; l--){ - segDistances[l]=segDistances[l-1]; - segQIndexes[l]=segQIndexes[l-1]; - segSIndexes[l]=segSIndexes[l-1]; - segIDs[l]=segIDs[l-1]; + for( l=trackNN-1 ; l > n ; l--){ + trackDistances[l]=trackDistances[l-1]; + trackQIndexes[l]=trackQIndexes[l-1]; + trackSIndexes[l]=trackSIndexes[l-1]; + trackIDs[l]=trackIDs[l-1]; } - segDistances[n]=thisDist; - segQIndexes[n]=qIndexes[0]; - segSIndexes[n]=sIndexes[0]; - successfulSegs++; - segIDs[n]=seg; + trackDistances[n]=thisDist; + trackQIndexes[n]=qIndexes[0]; + trackSIndexes[n]=sIndexes[0]; + successfulTracks++; + trackIDs[n]=track; break; } } @@ -1773,14 +1773,14 @@ } } // Duration match - // per-seg reset array values + // per-track reset array values for(unsigned k=0; k<pointNN; k++){ distances[k]=0.0; qIndexes[k]=~0; sIndexes[k]=~0; } - // Clean up current seg + // Clean up current track if(D!=NULL){ for(j=0; j<numVectors; j++) delete[] D[j]; @@ -1795,7 +1795,7 @@ gettimeofday(&tv2,NULL); if(verbosity>1) - cerr << endl << "processed segs :" << processedSegs << " matched segments: " << successfulSegs << " elapsed time:" + cerr << endl << "processed tracks :" << processedTracks << " matched tracks: " << successfulTracks << " elapsed time:" << ( tv2.tv_sec*1000 + tv2.tv_usec/1000 ) - ( tv1.tv_sec*1000+tv1.tv_usec/1000 ) << " msec" << endl; if(adbQueryResult==0){ @@ -1803,11 +1803,11 @@ cerr<<endl; // Output answer // Loop over nearest neighbours - for(k=0; k < min(segNN,successfulSegs); k++) - cout << fileTable+segIDs[k]*O2_FILETABLESIZE << " " << segDistances[k] << " " << segQIndexes[k] << " " << segSIndexes[k] << endl; + for(k=0; k < min(trackNN,successfulTracks); k++) + cout << fileTable+trackIDs[k]*O2_FILETABLESIZE << " " << trackDistances[k] << " " << trackQIndexes[k] << " " << trackSIndexes[k] << endl; } else{ // Process Web Services Query - int listLen = min(segNN, processedSegs); + int listLen = min(trackNN, processedTracks); adbQueryResult->__sizeRlist=listLen; adbQueryResult->__sizeDist=listLen; adbQueryResult->__sizeQpos=listLen; @@ -1818,17 +1818,17 @@ adbQueryResult->Spos = new int[listLen]; for(k=0; k<adbQueryResult->__sizeRlist; k++){ adbQueryResult->Rlist[k]=new char[O2_MAXFILESTR]; - adbQueryResult->Dist[k]=segDistances[k]; - adbQueryResult->Qpos[k]=segQIndexes[k]; - adbQueryResult->Spos[k]=segSIndexes[k]; - sprintf(adbQueryResult->Rlist[k], "%s", fileTable+segIDs[k]*O2_FILETABLESIZE); + adbQueryResult->Dist[k]=trackDistances[k]; + adbQueryResult->Qpos[k]=trackQIndexes[k]; + adbQueryResult->Spos[k]=trackSIndexes[k]; + sprintf(adbQueryResult->Rlist[k], "%s", fileTable+trackIDs[k]*O2_FILETABLESIZE); } } // Clean up - if(segOffsetTable) - delete segOffsetTable; + if(trackOffsetTable) + delete trackOffsetTable; if(queryCopy) delete queryCopy; //if(qNorm) @@ -1845,17 +1845,17 @@ } -// NBest matched filter distance between query and target segs +// NBest matched filter distance between query and target tracks // efficient implementation // outputs average of N minimum matched filter distances -void audioDB::segSequenceQueryEuc(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult){ +void audioDB::trackSequenceQueryEuc(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult){ initTables(dbName, inFile); // For each input vector, find the closest pointNN matching output vectors and report // we use stdout in this stub version unsigned numVectors = (statbuf.st_size-sizeof(int))/(sizeof(double)*dbH->dim); - unsigned numSegs = dbH->numFiles; + unsigned numTracks = dbH->numFiles; double* query = (double*)(indata+sizeof(int)); double* data = dataBuf; @@ -1874,6 +1874,7 @@ if(verbosity>1) cerr << "performing norms ... "; cerr.flush(); unsigned dbVectors = dbH->length/(sizeof(double)*dbH->dim); + // Make a copy of the query queryCopy = new double[numVectors*dbH->dim]; memcpy(queryCopy, query, numVectors*dbH->dim*sizeof(double)); @@ -1883,23 +1884,25 @@ assert(qNorm&&sNorm&&queryCopy&&sMeanL2&&sequenceLength); unitNorm(queryCopy, dbH->dim, numVectors, qNorm); query = queryCopy; + // Make norm measurements relative to sequenceLength unsigned w = sequenceLength-1; unsigned i,j; double* ps; double tmp1,tmp2; + // Copy the L2 norm values to core to avoid disk random access later on memcpy(sNorm, l2normTable, dbVectors*sizeof(double)); double* snPtr = sNorm; for(i=0; i<dbH->numFiles; i++){ - if(segTable[i]>=sequenceLength){ + if(trackTable[i]>=sequenceLength){ tmp1=*snPtr; j=1; w=sequenceLength-1; while(w--) *snPtr+=snPtr[j++]; ps = snPtr+1; - w=segTable[i]-sequenceLength; // +1 - 1 + w=trackTable[i]-sequenceLength; // +1 - 1 while(w--){ tmp2=*ps; *ps=*(ps-1)-tmp1+*(ps+sequenceLength-1); @@ -1907,13 +1910,13 @@ ps++; } ps = snPtr; - w=segTable[i]-sequenceLength+1; + w=trackTable[i]-sequenceLength+1; while(w--){ *ps=sqrt(*ps); ps++; } } - snPtr+=segTable[i]; + snPtr+=trackTable[i]; } double* pn = sMeanL2; @@ -1921,28 +1924,28 @@ while(w--) *pn++=0.0; ps=sNorm; - unsigned processedSegs=0; + unsigned processedTracks=0; for(i=0; i<dbH->numFiles; i++){ - if(segTable[i]>sequenceLength-1){ - w = segTable[i]-sequenceLength; + if(trackTable[i]>sequenceLength-1){ + w = trackTable[i]-sequenceLength; pn = sMeanL2+i; *pn=0; while(w--) if(*ps>0) *pn+=*ps++; - *pn/=segTable[i]-sequenceLength; + *pn/=trackTable[i]-sequenceLength; SILENCE_THRESH+=*pn; - processedSegs++; + processedTracks++; } - ps = sNorm + segTable[i]; + ps = sNorm + trackTable[i]; } if(verbosity>1) - cerr << "processedSegs: " << processedSegs << endl; + cerr << "processedTracks: " << processedTracks << endl; - SILENCE_THRESH/=processedSegs; + SILENCE_THRESH/=processedTracks; USE_THRESH=1; // Turn thresholding on - DIFF_THRESH=SILENCE_THRESH; // 50% of the mean shingle power + DIFF_THRESH=SILENCE_THRESH; // mean shingle power SILENCE_THRESH/=5; // 20% of the mean shingle power is SILENCE if(verbosity>4) cerr << "silence thresh: " << SILENCE_THRESH; @@ -1973,23 +1976,23 @@ if(verbosity>1) - cerr << "matching segs..." << endl; + cerr << "matching tracks..." << endl; assert(pointNN>0 && pointNN<=O2_MAXNN); - assert(segNN>0 && segNN<=O2_MAXNN); + assert(trackNN>0 && trackNN<=O2_MAXNN); // Make temporary dynamic memory for results - double segDistances[segNN]; - unsigned segIDs[segNN]; - unsigned segQIndexes[segNN]; - unsigned segSIndexes[segNN]; + double trackDistances[trackNN]; + unsigned trackIDs[trackNN]; + unsigned trackQIndexes[trackNN]; + unsigned trackSIndexes[trackNN]; double distances[pointNN]; unsigned qIndexes[pointNN]; unsigned sIndexes[pointNN]; - unsigned k,l,m,n,seg,segOffset=0, HOP_SIZE=sequenceHop, wL=sequenceLength; + unsigned k,l,m,n,track,trackOffset=0, HOP_SIZE=sequenceHop, wL=sequenceLength; double thisDist; double oneOverWL=1.0/wL; @@ -1999,11 +2002,11 @@ sIndexes[k]=~0; } - for(k=0; k<segNN; k++){ - segDistances[k]=0.0; - segQIndexes[k]=~0; - segSIndexes[k]=~0; - segIDs[k]=~0; + for(k=0; k<trackNN; k++){ + trackDistances[k]=0.0; + trackQIndexes[k]=~0; + trackSIndexes[k]=~0; + trackIDs[k]=~0; } // Timestamp and durations processing @@ -2035,7 +2038,7 @@ assert(meanDBdur); for(k=0; k<dbH->numFiles; k++){ meanDBdur[k]=0.0; - for(j=0; j<segTable[k]-1 ; j++) + for(j=0; j<trackTable[k]-1 ; j++) meanDBdur[k]+=timesTable[j+1]-timesTable[j]; meanDBdur[k]/=j; } @@ -2061,21 +2064,21 @@ assert(DD); gettimeofday(&tv1, NULL); - processedSegs=0; - unsigned successfulSegs=0; + processedTracks=0; + unsigned successfulTracks=0; double* qp; double* sp; double* dp; double diffL2; - // build segment offset table - unsigned *segOffsetTable = new unsigned[dbH->numFiles]; - unsigned cumSeg=0; - unsigned segIndexOffset; + // build track offset table + unsigned *trackOffsetTable = new unsigned[dbH->numFiles]; + unsigned cumTrack=0; + unsigned trackIndexOffset; for(k=0; k<dbH->numFiles;k++){ - segOffsetTable[k]=cumSeg; - cumSeg+=segTable[k]*dbH->dim; + trackOffsetTable[k]=cumTrack; + cumTrack+=trackTable[k]*dbH->dim; } char nextKey [MAXSTR]; @@ -2088,45 +2091,45 @@ double maxSample = 0; // Track loop - for(processedSegs=0, seg=0 ; processedSegs < dbH->numFiles ; seg++, processedSegs++){ + for(processedTracks=0, track=0 ; processedTracks < dbH->numFiles ; track++, processedTracks++){ - // get segID from file if using a control file - if(segFile){ - if(!segFile->eof()){ - segFile->getline(nextKey,MAXSTR); - seg=getKeyPos(nextKey); + // get trackID from file if using a control file + if(trackFile){ + if(!trackFile->eof()){ + trackFile->getline(nextKey,MAXSTR); + track=getKeyPos(nextKey); } else break; } - segOffset=segOffsetTable[seg]; // numDoubles offset - segIndexOffset=segOffset/dbH->dim; // numVectors offset + trackOffset=trackOffsetTable[track]; // numDoubles offset + trackIndexOffset=trackOffset/dbH->dim; // numVectors offset - if(sequenceLength<segTable[seg]){ // test for short sequences + if(sequenceLength<trackTable[track]){ // test for short sequences if(verbosity>7) - cerr << seg << "." << segIndexOffset << "." << segTable[seg] << " | ";cerr.flush(); + cerr << track << "." << trackIndexOffset << "." << trackTable[track] << " | ";cerr.flush(); // Sum products matrix for(j=0; j<numVectors;j++){ - D[j]=new double[segTable[seg]]; + D[j]=new double[trackTable[track]]; assert(D[j]); } // Matched filter matrix for(j=0; j<numVectors;j++){ - DD[j]=new double[segTable[seg]]; + DD[j]=new double[trackTable[track]]; assert(DD[j]); } double tmp; // Dot product for(j=0; j<numVectors; j++) - for(k=0; k<segTable[seg]; k++){ + for(k=0; k<trackTable[track]; k++){ qp=query+j*dbH->dim; - sp=dataBuf+segOffset+k*dbH->dim; + sp=dataBuf+trackOffset+k*dbH->dim; DD[j][k]=0.0; // Initialize matched filter array dp=&D[j][k]; // point to correlation cell j,k *dp=0.0; // initialize correlation cell @@ -2143,7 +2146,7 @@ for(j=0; j<numVectors-w; j++){ sp=DD[j]; spd=D[j+w]+w; - k=segTable[seg]-w; + k=trackTable[track]-w; while(k--) *sp+++=*spd++; } @@ -2154,7 +2157,7 @@ for(j=0; j<numVectors-w; j+=HOP_SIZE){ sp=DD[j]; spd=D[j+w]+w; - for(k=0; k<segTable[seg]-w; k+=HOP_SIZE){ + for(k=0; k<trackTable[track]-w; k+=HOP_SIZE){ *sp+=*spd; sp+=HOP_SIZE; spd+=HOP_SIZE; @@ -2163,13 +2166,13 @@ } if(verbosity>3 && usingTimes){ - cerr << "meanQdur=" << meanQdur << " meanDBdur=" << meanDBdur[seg] << endl; + cerr << "meanQdur=" << meanQdur << " meanDBdur=" << meanDBdur[track] << endl; cerr.flush(); } if(!usingTimes || (usingTimes - && fabs(meanDBdur[seg]-meanQdur)<meanQdur*timesTol)){ + && fabs(meanDBdur[track]-meanQdur)<meanQdur*timesTol)){ if(verbosity>3 && usingTimes){ cerr << "within duration tolerance." << endl; @@ -2178,10 +2181,10 @@ // Search for minimum distance by shingles (concatenated vectors) for(j=0;j<numVectors-wL;j+=HOP_SIZE) - for(k=0;k<segTable[seg]-wL;k+=HOP_SIZE){ - thisDist=2-(2/(qNorm[j]*sNorm[segIndexOffset+k]))*DD[j][k]; + for(k=0;k<trackTable[track]-wL;k+=HOP_SIZE){ + thisDist=2-(2/(qNorm[j]*sNorm[trackIndexOffset+k]))*DD[j][k]; if(verbosity>10) - cerr << thisDist << " " << qNorm[j] << " " << sNorm[segIndexOffset+k] << endl; + cerr << thisDist << " " << qNorm[j] << " " << sNorm[trackIndexOffset+k] << endl; // Gather chi^2 statistics if(thisDist<minSample) minSample=thisDist; @@ -2193,19 +2196,19 @@ logSampleSum+=log(thisDist); } - diffL2 = fabs(qNorm[j] - sNorm[segIndexOffset+k]); + // diffL2 = fabs(qNorm[j] - sNorm[trackIndexOffset+k]); // Power test if(!USE_THRESH || // Threshold on mean L2 of Q and S sequences - (USE_THRESH && qNorm[j]>SILENCE_THRESH && sNorm[segIndexOffset+k]>SILENCE_THRESH && + (USE_THRESH && qNorm[j]>SILENCE_THRESH && sNorm[trackIndexOffset+k]>SILENCE_THRESH && // Are both query and target windows above mean energy? - (qNorm[j]>qMeanL2*.25 && sNorm[segIndexOffset+k]>sMeanL2[seg]*.25))) // && diffL2 < DIFF_THRESH ))) + (qNorm[j]>qMeanL2*.25 && sNorm[trackIndexOffset+k]>sMeanL2[track]*.25))) // && diffL2 < DIFF_THRESH ))) thisDist=thisDist; // Computed above else thisDist=1000000.0; if(thisDist>=0 && thisDist<=radius){ distances[0]++; // increment count - break; // only need one seg point per query point + break; // only need one track point per query point } } // How many points were below threshold ? @@ -2213,25 +2216,25 @@ // Let's see the distances then... if(verbosity>3) - cerr << fileTable+seg*O2_FILETABLESIZE << " " << thisDist << endl; + cerr << fileTable+track*O2_FILETABLESIZE << " " << thisDist << endl; - // All the seg stuff goes here - n=segNN; + // All the track stuff goes here + n=trackNN; while(n--){ - if(thisDist>segDistances[n]){ - if((n==0 || thisDist<=segDistances[n-1])){ + if(thisDist>trackDistances[n]){ + if((n==0 || thisDist<=trackDistances[n-1])){ // Copy all values above up the queue - for( l=segNN-1 ; l > n ; l--){ - segDistances[l]=segDistances[l-1]; - segQIndexes[l]=segQIndexes[l-1]; - segSIndexes[l]=segSIndexes[l-1]; - segIDs[l]=segIDs[l-1]; + for( l=trackNN-1 ; l > n ; l--){ + trackDistances[l]=trackDistances[l-1]; + trackQIndexes[l]=trackQIndexes[l-1]; + trackSIndexes[l]=trackSIndexes[l-1]; + trackIDs[l]=trackIDs[l-1]; } - segDistances[n]=thisDist; - segQIndexes[n]=qIndexes[0]; - segSIndexes[n]=sIndexes[0]; - successfulSegs++; - segIDs[n]=seg; + trackDistances[n]=thisDist; + trackQIndexes[n]=qIndexes[0]; + trackSIndexes[n]=sIndexes[0]; + successfulTracks++; + trackIDs[n]=track; break; } } @@ -2240,7 +2243,7 @@ } } // Duration match - // Clean up current seg + // Clean up current track if(D!=NULL){ for(j=0; j<numVectors; j++) delete[] D[j]; @@ -2251,7 +2254,7 @@ delete[] DD[j]; } } - // per-seg reset array values + // per-track reset array values for(unsigned k=0; k<pointNN; k++){ distances[k]=0.0; qIndexes[k]=~0; @@ -2261,7 +2264,7 @@ gettimeofday(&tv2,NULL); if(verbosity>1){ - cerr << endl << "processed segs :" << processedSegs << " matched segments: " << successfulSegs << " elapsed time:" + cerr << endl << "processed tracks :" << processedTracks << " matched tracks: " << successfulTracks << " elapsed time:" << ( tv2.tv_sec*1000 + tv2.tv_usec/1000 ) - ( tv1.tv_sec*1000+tv1.tv_usec/1000 ) << " msec" << endl; cerr << "sampleCount: " << sampleCount << " sampleSum: " << sampleSum << " logSampleSum: " << logSampleSum << " minSample: " << minSample << " maxSample: " << maxSample << endl; @@ -2272,11 +2275,11 @@ cerr<<endl; // Output answer // Loop over nearest neighbours - for(k=0; k < min(segNN,successfulSegs); k++) - cout << fileTable+segIDs[k]*O2_FILETABLESIZE << " " << segDistances[k] << endl; + for(k=0; k < min(trackNN,successfulTracks); k++) + cout << fileTable+trackIDs[k]*O2_FILETABLESIZE << " " << trackDistances[k] << endl; } else{ // Process Web Services Query - int listLen = min(segNN, processedSegs); + int listLen = min(trackNN, processedTracks); adbQueryResult->__sizeRlist=listLen; adbQueryResult->__sizeDist=listLen; adbQueryResult->__sizeQpos=listLen; @@ -2287,17 +2290,17 @@ adbQueryResult->Spos = new int[listLen]; for(k=0; k<adbQueryResult->__sizeRlist; k++){ adbQueryResult->Rlist[k]=new char[O2_MAXFILESTR]; - adbQueryResult->Dist[k]=segDistances[k]; - adbQueryResult->Qpos[k]=segQIndexes[k]; - adbQueryResult->Spos[k]=segSIndexes[k]; - sprintf(adbQueryResult->Rlist[k], "%s", fileTable+segIDs[k]*O2_FILETABLESIZE); + adbQueryResult->Dist[k]=trackDistances[k]; + adbQueryResult->Qpos[k]=trackQIndexes[k]; + adbQueryResult->Spos[k]=trackSIndexes[k]; + sprintf(adbQueryResult->Rlist[k], "%s", fileTable+trackIDs[k]*O2_FILETABLESIZE); } } // Clean up - if(segOffsetTable) - delete[] segOffsetTable; + if(trackOffsetTable) + delete[] trackOffsetTable; if(queryCopy) delete[] queryCopy; //if(qNorm) @@ -2475,7 +2478,7 @@ // Literal translation of command line to web service -int adb__query(struct soap* soap, xsd__string dbName, xsd__string qKey, xsd__string keyList, xsd__string timesFileName, xsd__int qType, xsd__int qPos, xsd__int pointNN, xsd__int segNN, xsd__int seqLen, adb__queryResult &adbQueryResult){ +int adb__query(struct soap* soap, xsd__string dbName, xsd__string qKey, xsd__string keyList, xsd__string timesFileName, xsd__int qType, xsd__int qPos, xsd__int pointNN, xsd__int trackNN, xsd__int seqLen, adb__queryResult &adbQueryResult){ char queryType[256]; for(int k=0; k<256; k++) queryType[k]='\0'; @@ -2483,15 +2486,15 @@ strncpy(queryType, "point", strlen("point")); else if (qType == O2_FLAG_SEQUENCE_QUERY) strncpy(queryType, "sequence", strlen("sequence")); - else if(qType == O2_FLAG_SEG_QUERY) - strncpy(queryType,"segment", strlen("segment")); + else if(qType == O2_FLAG_TRACK_QUERY) + strncpy(queryType,"track", strlen("track")); else strncpy(queryType, "", strlen("")); if(pointNN==0) pointNN=10; - if(segNN==0) - segNN=10; + if(trackNN==0) + trackNN=10; if(seqLen==0) seqLen=16; @@ -2499,8 +2502,8 @@ sprintf(qPosStr, "%d", qPos); char pointNNStr[256]; sprintf(pointNNStr,"%d",pointNN); - char segNNStr[256]; - sprintf(segNNStr,"%d",segNN); + char trackNNStr[256]; + sprintf(trackNNStr,"%d",trackNN); char seqLenStr[256]; sprintf(seqLenStr,"%d",seqLen); @@ -2520,8 +2523,8 @@ qPosStr, COM_POINTNN, pointNNStr, - COM_SEGNN, - segNNStr, // Need to pass a parameter + COM_TRACKNN, + trackNNStr, // Need to pass a parameter COM_SEQLEN, seqLenStr };
--- a/audioDB.h Fri Aug 10 04:52:33 2007 +0000 +++ b/audioDB.h Mon Aug 13 19:14:33 2007 +0000 @@ -16,7 +16,7 @@ -d, --database=filename database name to be used with database commands -N, --new make a new database -S, --status database information - -D, --dump list all segments: index key size + -D, --dump list all tracks: index key size Database Insertion: The following commands process a binary input feature file and optional @@ -46,12 +46,12 @@ database using the named feature vector file as a query -q, --qtype=type the type of search (possible values="point", - "segment", "sequence" default=`sequence') + "track", "sequence" default=`sequence') -p, --qpoint=position ordinal position of query vector (or start of sequence) in feature vector input file (default=`0') -n, --pointnn=numpoints number of point nearest neighbours to use [per - segment in segment and sequence mode] + track in track and sequence mode] (default=`10') -r, --resultlength=length maximum length of the result list (default=`10') @@ -107,7 +107,7 @@ #define COM_SEQLEN "--sequencelength" #define COM_SEQHOP "--sequencehop" #define COM_POINTNN "--pointnn" -#define COM_SEGNN "--resultlength" +#define COM_TRACKNN "--resultlength" #define COM_QPOINT "--qpoint" #define COM_FEATURES "--features" #define COM_QUERYKEY "--key" @@ -117,7 +117,7 @@ #define O2_MAGIC 1111765583 // 'B'<<24|'D'<<16|'2'<<8|'O' reads O2DB in little endian order #define O2_DEFAULT_POINTNN (10U) -#define O2_DEFAULT_SEGNN (10U) +#define O2_DEFAULT_TRACKNN (10U) #define O2_DEFAULTDBSIZE (2000000000) // 2GB table size //#define O2_DEFAULTDBSIZE (1000000000U) // 1GB table size @@ -126,7 +126,7 @@ #define O2_MAXFILES (10000U) // 10,000 files #define O2_MAXFILESTR (256U) #define O2_FILETABLESIZE (O2_MAXFILESTR) -#define O2_SEGTABLESIZE (sizeof(unsigned)) +#define O2_TRACKTABLESIZE (sizeof(unsigned)) #define O2_HEADERSIZE (sizeof(dbTableHeaderT)) #define O2_MEANNUMVECTORS (1000U) #define O2_MAXDIM (1000U) @@ -137,7 +137,7 @@ #define O2_FLAG_MINMAX (0x2U) #define O2_FLAG_POINT_QUERY (0x4U) #define O2_FLAG_SEQUENCE_QUERY (0x8U) -#define O2_FLAG_SEG_QUERY (0x10U) +#define O2_FLAG_TRACK_QUERY (0x10U) #define O2_FLAG_TIMES (0x20U) // Error Codes @@ -167,8 +167,8 @@ const char *inFile; const char *hostport; const char *key; - const char* segFileName; - ifstream *segFile; + const char* trackFileName; + ifstream *trackFile; const char *command; const char *timesFileName; ifstream *timesFile; @@ -180,13 +180,13 @@ struct stat statbuf; dbTableHeaderPtr dbH; size_t fileTableOffset; - size_t segTableOffset; + size_t trackTableOffset; size_t dataoffset; size_t l2normTableOffset; size_t timesTableOffset; char *fileTable; - unsigned* segTable; + unsigned* trackTable; double* dataBuf; double* inBuf; double* l2normTable; @@ -198,7 +198,7 @@ unsigned verbosity; // how much do we want to know? unsigned queryType; // point queries default unsigned pointNN; // how many point NNs ? - unsigned segNN; // how many seg NNs ? + unsigned trackNN; // how many track NNs ? unsigned sequenceLength; unsigned sequenceHop; unsigned queryPoint; @@ -218,9 +218,9 @@ void error(const char* a, const char* b = ""); void pointQuery(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult=0); void sequenceQuery(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult=0); - void segPointQuery(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult=0); - void segSequenceQuery(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult=0); - void segSequenceQueryEuc(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult=0); + void trackPointQuery(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult=0); + void trackSequenceQuery(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult=0); + void trackSequenceQueryEuc(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult=0); void initTables(const char* dbName, const char* inFile); void NBestMatchedFilter(); @@ -242,7 +242,7 @@ void query(const char* dbName, const char* inFile, adb__queryResult *adbQueryResult=0); void status(const char* dbName); void ws_status(const char*dbName, char* hostport); - void ws_query(const char*dbName, const char *segKey, const char* hostport); + void ws_query(const char*dbName, const char *trackKey, const char* hostport); void l2norm(const char* dbName); void dump(const char* dbName); void deleteDB(const char* dbName, const char* inFile);
--- a/gengetopt.in Fri Aug 10 04:52:33 2007 +0000 +++ b/gengetopt.in Mon Aug 13 19:14:33 2007 +0000 @@ -19,13 +19,13 @@ # -T --timePointsList.txt # -B --BATCHINSERT dbName # -# -Q --QUERY {point|segment|sequence} +# -Q --QUERY {point|track|sequence} # -p --qpoint - ordinal position of query vector in feature input file -# -n --pointnn n - number of nearest neigbour points per segment -# -r --resultlength n - number of segments (nearest neighbours) to return +# -n --pointnn n - number of nearest neigbour points per track +# -r --resultlength n - number of tracks (nearest neighbours) to return # -l --seqlen len - length of sequence # -h --seqhop hop - hop size of sequence -# -R --radius - radius-based search, return all points/segments/sequences < radius (0...Inf) +# -R --radius - radius-based search, return all points/tracks/sequences < radius (0...Inf) # -x --x - time expand (compress) factor (ratio of result length to query length (and visa-versa)) [1..Inf] # -o --rotate - rotate the query feature vectors on search # @@ -54,18 +54,18 @@ option "key" k "unique identifier associated with features." string typestr="identifier" dependon="features" optional text "" option "BATCHINSERT" B "add feature vectors named in a --featureList file (with optional keys in a --keyList file) to the named database." dependon="featureList" optional -option "featureList" F "text file containing list of binary feature vector files to process" string typestr="filename" dependon="database" optional +option "featureList" F "text file containing list of binary feature vector files to process, one per track" string typestr="filename" dependon="database" optional option "timesList" T "text file containing list of ascii --times for each --features file in --featureList." string typestr="filename" dependon="featureList" optional option "keyList" K "text file containing list of unique identifiers associated with --features." string typestr="filename" optional section "Database Search" sectiondesc="Thse commands control the retrieval behaviour.\n" -option "QUERY" Q "content-based search on --database using --features as a query. Optionally restrict the search to those segments identified in a --keyList." values="point","segment","sequence" typestr="searchtype" dependon="database" dependon="features" optional +option "QUERY" Q "content-based search on --database using --features as a query. Optionally restrict the search to those tracks identified in a --keyList." values="point","track","sequence" typestr="searchtype" dependon="database" dependon="features" optional option "qpoint" p "ordinal position of query start point in --features file." int typestr="position" default="0" optional option "exhaustive" e "exhaustive search: iterate through all query vectors in search. Overrides --qpoint." flag off optional hidden option "pointnn" n "number of point nearest neighbours to use in retrieval." int typestr="numpoints" default="10" optional -option "radius" R "radius search, returns all points/segments/sequences inside given radius." double default="1.0" optional hidden +option "radius" R "radius search, returns all points/tracks/sequences inside given radius." double default="1.0" optional hidden option "expandfactor" x "time compress/expand factor of result length to query length [1.0 .. 100.0]." double default="1.1" optional hidden option "rotate" o "rotate query vectors for rotationally invariant search." flag off optional hidden option "resultlength" r "maximum length of the result list." int typestr="length" default="10" optional