Changeset 2606 for Trunk/Scripts/Audio_scripts/AudioSegmentation/AudioBook/Audio.pm

Timestamp:

06/09/08 17:06:50 (6 months ago)

Author:

kmaclean

Message:

refacture to create Chapter, Segments & MissingWords? classes - snapshop

Files:

• Trunk/Scripts/Audio_scripts/AudioSegmentation/AudioBook/Audio.pm (modified) (22 diffs)

Trunk/Scripts/Audio_scripts/AudioSegmentation/AudioBook/Audio.pm

@@ -1 +1 @@
-#! /usr/bin/perl
-
+.1
-
-=head1 NAME
-
-Book::Audio
+
-
-=cut 
@@ -14 +14 @@
-use File::Basename;
-use Audio::Wav;
-
+
-use AudioBook::Text;
+
-####################################################################
-### Class Variables
-####################################################################
-found, $max_sentence_length_
+
-my $min_sentence_length_found = 0;
-my $min_sentence_length_linenumber = 0;
+my $max_sentence_length_found=0;
-my $up_increment = 1;
-my $down_increment = -1;
@@ -36 +38 @@
-sub new {
-        my ($class,$super) = @_; 
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
-}    
-
@@ -82 +82 @@
-}
-
-_
+
-
-creates an MLF (HTK required "Multi-Label File") file for a prompt.
@@ -101 +101 @@
-=head1 INSTANCE METHODS
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-Public Method that does the actual segmentation of the speech audio file using the text file supplied by the user. 
@@ -110 +262 @@
-=cut
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
-        my $filename_nopath = basename($filename);
-        my $filename_nosuffix = fileparse($filename, "wav");
@@ -122 +278 @@
-        $self->{"filename_nopath"} = $filename_nopath;
-        $self->{"filename_prefix"} = lc(substr($filename_nopath,0,3));
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
-                
-
-
+
+
+
-        # copy to "interim_files"" directory for processing; also converts to 16 bits per sample (-w=16-bits) so it can be processed by HVite
-
+ if $debug
-        
-        my $loop;
@@ -204 +353 @@
-}
-
-=head3 _getSampleRate 
-
-gets the sample rate of the audio file using Audio::Wav module
-
-=cut
-
-sub _getSampleRate { # private
-        my ($self) = @_;        
-        my $debug = $self->{"debug"};
-        my $wavfilename = $self->{"filename"};
-        
-        my $wav = new Audio::Wav  || confess "AUDIO.pm error: $?";
-        my $read = $wav -> read( "$wavfilename" ) || confess "AUDIO.pm error: $?";
-    my $details = $read -> details() || confess "AudioBook.pl error: $?";
-        return ($$details{sample_rate}); #sampling rate
-}
-
-=head3 _forceAlign 
-
-Performs "Forced Alignement" of the user submitted speech audio file.  This is the process of taking the text transcription 
-of the speech audio file and figuring you the time stamps for each of the words.  This helps determine where the pauses are.
-
-Sox audio editor is used to downsample the audio to 16kHz-16bit - Hvite only works with 16kHz sampling rate audio
-
-The time stamps are put in the "AudioBooks/interim_files/aligned.out" file 
-
-=cut
-
-sub     _forceAlign { # private
-        my ($self) = @_;        
-        my $debug = $self->{"debug"};   
-        my $filename = $self->{"filename"};
-        my $htk_files = $self->{'htk_files'};
-        my $textContents = $self->{"textContents"}; 
-        my $beam_width = $self->{"beam_width"};
-        
-        # Hvite only works with 16kHz sampling rate audio
-        $command = ("sox $filename -c 1 -r 16000 -w AudioBook/interim_files/downsampled.wav"); print "$command\n"; system($command) == 0 or confess "fullrun $command failed: $?";
-        # create mlf file 
-        $textContents->createMLFFile("downsampled","AudioBook/interim_files/words.mlf" );
-        # forced alignment - creates aligned.out
-        print "\nRunning HVite ...\n";
-        print "check interim_files/logs/HVite_log for a possible errors\n"; 
-        print "(like \"no tokens surviving\"... which means that text does not match audio)\n\n";
-        ####################################################################    
-        # process audio file with HVite without using a script file (i.e. -S $htk_files/train.scp); need to remove -l '*'
-        #HVite -A -D -T 1  -a -b SENT-END -m -C AudioBook/input_files/htk/wav_config 
-        #   -H AudioBook/input_files/htk/models/macros 
-        #   -H AudioBook/input_files/htk/models/hmmdefs 
-        #   -m -t 250.0 150.0 1000.0 -I AudioBook/interim_files/words.mlf 
-        #   -i AudioBook/interim_files/aligned.out 
-        #   AudioBook/interim_files/dict AudioBook/input_files/htk/models/tiedlist  
-        #   AudioBook/interim_files/downsampled.wav
-        # !!!!!!
-#       $command = ("pwd && HVite -A -D -T 1 -l '*' -a -b SENT-END -m -C $htk_files/wav_config -H $htk_files/models/macros -H $htk_files/models/hmmdefs -m -t 250.0 150.0 1000.0 -I AudioBook/interim_files/words.mlf  -i AudioBook/interim_files/aligned.out -S $htk_files/train.scp AudioBook/interim_files/dict $htk_files/models/tiedlist > AudioBook/interim_files/logs/HVite_log"); system($command) == 0 or confess "error: $command failed: $?";
-        $command = ("pwd && HVite -A -D -T 1 -l '*' -a -b SENT-END -m -C $htk_files/wav_config -H $htk_files/models/macros -H $htk_files/models/hmmdefs -m -t 500.0 150.0 8000.0 -I AudioBook/interim_files/words.mlf  -i AudioBook/interim_files/aligned.out -S $htk_files/train.scp AudioBook/interim_files/dict $htk_files/models/tiedlist > AudioBook/interim_files/logs/HVite_log"); system($command) == 0 or confess "error: $command failed: $?";
-# !!!!!!
-        open (HVite_Log,"AudioBook/interim_files/logs/HVite_log") || confess "error: can't open AudioBook/interim_files/HVite_log: $?";
-        while (my $line = <HVite_Log>) {
-                chomp $line;
-                my $filename;
-                if ($line =~ /Aligning File:/) { 
-                        my @line=split(/:/, $line);
-                        $filename = pop(@line);
-                } elsif ($line =~ /No tokens survived to final node of network at beam/) {
-                        my @line=split(/ /, $line);
-                        my $beam = pop (@line);
-                        $beam =~ s/ //g;
-                        print "\n\n******************************************************************\n";
-                        print "**** check that audio corresponds to prompt in file ***\n";
-                        print "******************************************************************\n\n";
-                        if ($beam > $beam_width) {
-                                confess "audio not corresponding to prompt file, check HVite_Log; error code: $?" ;
-                        }       
-                } 
-        }
-}
-
-=head3 _processHViteOutput 
-
-This method reads the HVite output in the "aligned.out" file (generated by the _forceAlign method) and parses it into a 
-more user friendly format, and puts into the "AudioBooks/interim_files/htksegment_log" file. 
-
-=cut
-
-sub _processHViteOutput { # private
-        my ($self) = @_;        
-        my $debug = $self->{"debug"};   
-                
-        open(ALIGNED, "<AudioBook/interim_files/aligned.out") or confess ("can\'t open AudioBook/interim_files/aligned.out file for reading"); 
-        my @aligned = <ALIGNED>;
-        close ALIGNED;
-        my ($pause, @aligned_words, $current_word, $current_startTime, $current_endTime);
-        my $first_pass = 1;
-        foreach my $line (@aligned) {
-                chomp $line;
-                my @labels = split(/ /, $line); 
-                #print "@labels\n";
-                my ($startTime,$endTime,$phone,$score,$word)= @labels;  
-                if (!defined($word)) {$word=""};
-                if (!defined($phone)) {$phone=""};
-                if ($phone eq "sp") {
-                        $pause = $endTime - $startTime;
-                }
-                if (($word ne "") and ($word ne "SENT-END") and ($first_pass)) {        
-                        $current_word = $word;          
-                        $current_startTime = $startTime;
-                        $first_pass = undef;                                    
-                } elsif (($word ne "") and (!$first_pass)) {    
-                        #print  "$startTime,$endTime,$phone,$score,$word,pause:$pause\n";
-                        $current_endTime = $startTime-$pause;
-                        push (@aligned_words, "$current_word $current_startTime $current_endTime $pause");      
-                        $current_word = $word;          
-                        $current_startTime = $startTime;                        
-                }       
-        }
-        open(LOG, ">AudioBook/interim_files/htksegment_log") or confess ("cannot open htksegment_log for output");      
-        foreach my $line (@aligned_words) {
-                print LOG "$line\n";
-        }
-        close (LOG);
-
-        return (\@aligned_words);
-}
-
-=head3 _sentence_test 
-
@@ -338 +362 @@
-sub _sentence_test { # private
-        my ($self, $where, $increment,$aligned_words, $fileid, $sentence_start, $sentence_end) = @_;
-
+
+
+
+
-        my $filename_prefix = $self->{"filename_prefix"};
-        my $average_sentence_length = $self->{"average_sentence_length"};
-        #my $max_sentence_length = $self->{"max_sentence_length"};
-        my $min_pause_for_sentence_break = $self->{"min_pause_for_sentence_break"};
-        
-        my @aligned_line = split(/\s/,$$aligned_words[$$sentence_end + $increment]);
@@ -415 +439 @@
-sub _calculateStats { # Calculate min and max sentence
-        my ($self,$sentence_start,$sentence_end,$increment,$filename_prefix,$padded_fileid ) = @_;
-
-
-
-
+
+
+
-        
-        if ($debug) {
@@ -433 +456 @@
-                push (@max_sentences,"$filename_prefix$padded_fileid:" . ($sentence_length+1));
-        } 
+        #print "!!!!!!!sentence_length:$sentence_length;min_sentence_length_found:$min_sentence_length_found\n";
+        #print "!!!!!!!sentence_length:$sentence_length>max_sentence_length_found:$max_sentence_length_found\n";
-        if ($sentence_length<$min_sentence_length_found) {
-        $min_sentence_length_found = $sentence_length+1; 
@@ -450 +475 @@
-sub _last_sentence { # private
-        my ($self, $aligned_words, $fileid, $sentence_start, $sentence_end) = @_;
-
+
+
-        my $filename_prefix = $self->{"filename_prefix"};
-        my ($count, $startTime);
@@ -489 +515 @@
-sub _processAudio { # private
-        my ($self, $startTime, $endTime,$padded_fileid) = @_;
-
+
+
-        my $process_audio = $self->{"process_audio"};
-        my $samplerate = $self->{"samplerate"};
-        my $filename_prefix = $self->{"filename_prefix"};
-        my $filename_nopath = $self->{"filename_nopath"};
-self->{'htk_files'}
+audioBook->getHtk_files()
-
-        #print "debug:$debug;process_audio:$process_audio:\n";
@@ -513 +540 @@
-=head2 verifySegments 
-
-Create a segmented audio file from the original audio file using startTime and endTime generated from segmentation routines.
-
-Second pass Forced Alignment on the segmented audio files generated from the segment method.  It tries to perform forced aligment 
-y match.  If they do not, it flags 
-the prompt for further editing
+ text matches the what was said
+in the audio file.  If they do not match, it flags the promptID, so that you can manually correct any errors
- 
-=cut
@@ -523 +548 @@
-sub verifySegments { #public
-        my ($self) = @_;                
-
-
+
+
+
-        
-        open(LOG,">>$log") or confess ("cannot open $log file");
@@ -533 +559 @@
-        print "-------------------------------------------------------------\n";                
-        open(PROMPTS, "<AudioBook/interim_files/prompts") or confess ("cannot open AudioBook/output_files/prompts for output");
+        my @promptsWithTimes;
-        my @promptScores;       
-        foreach my $line (<PROMPTS>) {
@@ -544 +571 @@
-                my $avgLogLikelihoodPerFrame = $self->_processHviteLog($wavfilename);
-                push (@promptScores,[$avgLogLikelihoodPerFrame,$fileID,"@promptList"]);
+                # !!!!!!
+                push (@promptsWithTimes,[$fileID,$self->_processHViteOutput()]); # converts aligned.out to htksegment_log
+                # !!!!!!
-                print "\.";
-        }
-        print "\n";
-        close (PROMPTS);                
-
+#
-        #my @sortedlist = sort @promptScores;
-        my @sortedlist = sort { $a->[0] cmp $b->[0] } (@promptScores); # sort of 1st element of @promptScores (which is an array of arrays)
@@ -563 +593 @@
-        }
-        close LOG;
+        
+        $self->{"promptsWithScores"} = \@promptScores;
+        return (\@promptsWithTimes);
-}
-
@@ -573 +606 @@
-sub _createMLFFile { # private
-        my ($self,$promptList)= @_;
-
+
+
-        open(MLF, ">AudioBook/interim_files/segment.mlf") or confess ("cannot open AudioBook/interim_files/segment.mlf for output");    
-        print MLF "#!MLF!#\n"; # 
@@ -591 +625 @@
-sub _forceAlignSegment {  # private
-        my ($self,$wavfilename, $segmentMLF)= @_;
-
-
+
+
+
-
-        $command = ("sox AudioBook/interim_files/wav/$wavfilename -c 1 -r 16000 -w AudioBook/interim_files/downsampled.wav 2>&1 > AudioBook/interim_files/logs/Segment_Soxlog"); system($command) == 0 or confess "$command failed: $?";
@@ -608 +643 @@
-sub _processHviteLog {  # private
-        my ($self,$wavfilename)= @_;
-
-
+
+
+
-        my $avgLogLikelihoodPerFrame;
-        my $log = $self->{"log"};       
@@ -672 +708 @@
-=head1 Change Log    
-
-
-
-
+
+
+
+
-        
-=head1 Current Unresolved Issues: