Improved cli app, small fixes

CLI app can now accept flac, speex or wav files as input. Small fixes in other scripts.
zaf · Mar 4, 2012 · 9d10199 · 9d10199
1 parent 8dd27fc
commit 9d10199
Show file tree

Hide file tree

Showing 3 changed files with 115 additions and 39 deletions.
diff --git a/samples/speech-recog-cli.pl b/samples/speech-recog-cli.pl
@@ -9,7 +9,7 @@
 # the GNU General Public License Version 2. See the COPYING file
 # at the top of the source tree.
 #
-# The script takes as input flac files at 8kHz and returns the following values:
+# The script takes as input flac, speex or wav files at 8kHz and returns the following values:
 # status     : Return status. 0 means success, non zero values indicating different errors.
 # id         : Some id string that googles engine returns, not very useful(?).
 # utterance  : The generated text string.
@@ -20,24 +20,55 @@
 
 use strict;
 use warnings;
+use File::Temp qw(tempfile);
+use Getopt::Std;
+use File::Basename;
 use LWP::UserAgent;
 
-if (!@ARGV || $ARGV[0] eq '-h' || $ARGV[0] eq '--help') {
+my %options;
-	print "Speech recognition using google voice.\n\n";
+my $filetype;
-	print "Usage: $0 [FILES]\n\n";
-	exit;
-}
-
 my $url        = "http://www.google.com/speech-api/v1/recognize?xjerr=1&client=chromium";
 my $samplerate = 8000;
-#my $filetype   = "x-speex-with-header-byte";
-my $filetype   = "x-flac";
 my $language   = "en-US";
 my $results    = 1;
+
+getopts('l:r:hq', \%options);
+
+VERSION_MESSAGE() if (defined $options{h} || !@ARGV);
+
+if (defined $options{l}) {
+# check if language setting is valid #
+	if ($options{l} =~ /^[a-z]{2}(-[a-zA-Z]{2,6})?$/) {
+		$language = $options{l};
+	} else {
+		say_msg("Invalid language setting. Using default.");
+	}
+}
+
+if (defined $options{r}) {
+# set number or results #
+	$results = $options{r} if ($options{r} =~ /%d+/);
+}
+
 my @file_list  = @ARGV;
 
 foreach my $file (@file_list) {
-	print "Openning $file\n";
+	my($filename, $dir, $ext) = fileparse($file, qr/\.[^.]*/);
+	if ($ext ne ".flac" && $ext ne ".spx" && $ext ne ".wav") {
+		say_msg("Unsupported filetype: $ext");
+		exit 1;
+	}
+
+	if ($ext eq ".flac") {
+		$filetype   = "x-flac";
+	} elsif ($ext eq ".spx") {
+		$filetype   = "x-speex-with-header-byte";
+	} elsif ($ext eq ".wav") {
+		$filetype   = "x-flac";
+		$file = encode_flac($file);
+	}
+
+	say_msg("Openning $file");
 	open(my $fh, "<", "$file") or die "Cant read file: $!";
 	my $audio = do { local $/; <$fh> };
 	close($fh);
@@ -54,15 +85,60 @@
 	if ($response->content =~ /^\{"status":(\d*),"id":"(.*)","hypotheses":\[(.*)\]\}$/) {
 		$response{status} = "$1";
 		$response{id}     = "$2";
-		if ($response{status} == 5) {
+		if ($response{status} != 0) {
-			print "Error reading audio file\n";
+			say_msg("Error reading audio file");
 		}
 		if ($3 =~ /^\{"utterance":"(.*)","confidence":(.*)\}/) {
 			$response{utterance}  = "$1";
 			$response{confidence} = "$2";
 		}
 	}
-	printf "%-10s : %s\n", $_, $response{$_} foreach (keys %response);
+	if (!defined $options{q}) {
-	#print $response->content;
+		printf "%-10s : %s\n", $_, $response{$_} foreach (keys %response);
+	} else {
+		print "$response{utterance}\n";
+	}
+	print $response->content;
+}
+exit 0;
+
+sub encode_flac {
+# Encode file to flac and return the filename #
+	my $file   = shift;
+	my $tmpdir = "/tmp";
+	my $flac   = `/usr/bin/which flac`;
+
+	if (!$flac) {
+		say_msg("flac encoder is missing. Aborting.");
+		exit 1;
+	}
+	chomp($flac);
+
+	my ($fh, $tmpname) = tempfile("recg_XXXXXX",
+		DIR => $tmpdir,
+		SUFFIX => '.flac',
+		UNLINK => 1,
+	);
+	if (system($flac, "-8", "-f", "--totally-silent", "-o", "$tmpname", "$file")) {
+		say_msg("$flac failed to encode file.");
+		exit 1;
+	}
+	return $tmpname;
+}
+
+sub say_msg {
+# Print messages to user if 'quiet' flag is not set #
+	print @_, "\n" if (!defined $options{q});
+}
+
+sub VERSION_MESSAGE {
+# Help message #
+	print "Speech recognition using google voice API.\n\n",
+		 "Usage: $0 [options] [file(s)]\n\n",
+		 "Supported options:\n",
+		 " -l <lang>      specify the language to use, defaults to 'en-US' (English)\n",
+		 " -r <number>    specify the number of results\n",
+		 " -q             Return only recognised utterance. Don't print any messages or warnings\n",
+		 " -h             this help message\n\n";
+	exit 1;
 }
-exit;
diff --git a/samples/wolfram/wolfram.agi b/samples/wolfram/wolfram.agi
@@ -63,12 +63,12 @@ while (<STDIN>) {
 $name = " -- $AGI{request}:";
 
 if (!$app_id) {
-	&return_var("You must have an Application ID from WolframAlpha to use this script.");
+	return_var("You must have an Application ID from WolframAlpha to use this script.");
 	die "$name You must have an Application ID from WolframAlpha to use this script.\n";
 }
 
 if (!length($AGI{arg_1})) {
-	&return_var("No text passed to Wolfram.");
+	return_var("No text passed to Wolfram.");
 	die "$name No text passed, Aborting.\n";
 }
 print STDERR "$name Questions is: $AGI{arg_1}\n" if($debug);
@@ -90,40 +90,40 @@ my $ua_request = HTTP::Request->new(
 );
 my $ua_response = $ua->request($ua_request);
 if (!$ua_response->is_success) {
-	&return_var("Failed to contact Wolfram server.");
+	return_var("Failed to contact Wolfram server.");
 	die "$name Failed to contact server.";
 }
 
 $w_reply = XMLin($ua_response->content);
 if ($w_reply->{success} eq 'false') {
 	print STDERR "$name Wolfram returned no answer.\n" if ($debug);
-	&return_var("I don't know how to answer that.");
+	return_var("I don't know how to answer that.");
 	exit;
 }
 
 # Trying to locate answers in predifined fields in Wolframs output. #
 foreach (keys %{$w_reply->{pod}}) {
 	if (/subpod/) {
-		&add_answer("$w_reply->{pod}{$_}{plaintext}\n");
+		add_answer("$w_reply->{pod}{$_}{plaintext}\n");
 		$results++;
 		last;
 	} elsif (/Result|Value/) {
-		eval{ &add_answer("$w_reply->{pod}{$_}{subpod}{plaintext}\n"); };
+		eval{ add_answer("$w_reply->{pod}{$_}{subpod}{plaintext}\n"); };
-		eval{ &add_answer("$w_reply->{pod}{$_}{subpod}[0]{plaintext}\n"); };
+		eval{ add_answer("$w_reply->{pod}{$_}{subpod}[0]{plaintext}\n"); };
-		eval{ &add_answer("$w_reply->{pod}{$_}{subpod}[1]{plaintext}\n"); };
+		eval{ add_answer("$w_reply->{pod}{$_}{subpod}[1]{plaintext}\n"); };
 		$results++;
 		last;
 	} elsif (/Definition:WordData|Basic:ChemicalData|ComparisonAsLength|Comparison|
 				BasicInformation|NotableFacts:PeopleData|Basic|Properties/x
 			) {
-		&add_answer("$w_reply->{pod}{$_}{subpod}{plaintext}\n");
+		add_answer("$w_reply->{pod}{$_}{subpod}{plaintext}\n");
 		$results++;
 	} elsif (/WeatherForecast:WeatherData/) {
-		&add_answer("$w_reply->{pod}{$_}{subpod}[0]{title} ",
+		add_answer("$w_reply->{pod}{$_}{subpod}[0]{title} ",
 			"$w_reply->{pod}{$_}{subpod}[0]{plaintext}\n");
-		&add_answer("$w_reply->{pod}{$_}{subpod}[1]{title} ",
+		add_answer("$w_reply->{pod}{$_}{subpod}[1]{title} ",
 			"$w_reply->{pod}{$_}{subpod}[1]{plaintext}\n");
-		eval{ &add_answer("$w_reply->{pod}{$_}{subpod}[2]{title} ",
+		eval{ add_answer("$w_reply->{pod}{$_}{subpod}[2]{title} ",
 			"$w_reply->{pod}{$_}{subpod}[2]{plaintext}\n"); };
 		$results++;
 		last;
@@ -133,20 +133,20 @@ foreach (keys %{$w_reply->{pod}}) {
 # If the above did't work search for answers all available fields. #
 if (!$results) {
 	foreach (keys %{$w_reply->{pod}}) {
-		eval{ &add_answer("$_: $w_reply->{pod}{$_}{subpod}{plaintext}\n"); };
+		eval{ add_answer("$_: $w_reply->{pod}{$_}{subpod}{plaintext}\n"); };
-		eval{ &add_answer("$_: $w_reply->{pod}{$_}{subpod}[0]{plaintext}\n"); };
+		eval{ add_answer("$_: $w_reply->{pod}{$_}{subpod}[0]{plaintext}\n"); };
-		eval{ &add_answer("$_: $w_reply->{pod}{$_}{subpod}[1]{plaintext}\n"); };
+		eval{ add_answer("$_: $w_reply->{pod}{$_}{subpod}[1]{plaintext}\n"); };
 		$results++;
 	}
 }
 
 # In case we were not able to locate any answers. #
 if (!$results) {
 	print STDERR "$name Failed to get any answer.\n" if ($debug);
-	&return_var("Failed to get any answer.");
+	return_var("Failed to get any answer.");
 } else {
 	print STDERR "$name Answer is:$answer\n" if ($debug);
-	&return_var($answer);
+	return_var($answer);
 }
 exit;
 
@@ -169,7 +169,7 @@ sub return_var {
 	my $text = shift;
 	print STDERR "$name Setting $var_name to $text\n" if ($debug);
 	print "SET VARIABLE \"$var_name\" \"$text\"\n";
-	&checkresponse();
+	checkresponse();
 }
 
 sub checkresponse {

diff --git a/speech-recog.agi b/speech-recog.agi
@@ -169,11 +169,11 @@ if (length($AGI{arg_4})) {
 # Answer channel if not already answered #
 print STDERR "$name Checking channel status.\n" if ($debug);
 print "CHANNEL STATUS\n";
-@result = &checkresponse();
+@result = checkresponse();
 if ($result[0] == 4) {
 	print STDERR "$name Answering channel.\n" if ($debug);
 	print "ANSWER\n";
-	&checkresponse();
+	checkresponse();
 }
 
 # Reset variables. #
@@ -186,7 +186,7 @@ if ($result[0] == 4) {
 print STDERR "$name Clearing channel variables.\n" if ($debug);
 foreach (keys %response) {
 	print "SET VARIABLE \"$_\" \"$response{$_}\"\n";
-	&checkresponse();
+	checkresponse();
 }
 
 # Hnadle interrupts #
@@ -196,7 +196,7 @@ $SIG{'HUP'} = \&int_handler;
 # Record file #
 ($fh, $tmpname) = tempfile("stt_XXXXXX", DIR => $tmpdir, UNLINK => 1);
 print "RECORD FILE $tmpname $format \"$intkey\" \"-1\" $beep \"$silence\"\n";
-@result = &checkresponse();
+@result = checkresponse();
 die "$name Failed to record file, aborting...\n" if ($result[0] == -1);
 
 if ($debug) {
@@ -206,7 +206,7 @@ if ($debug) {
 		"$silence, Interrupt keys: $intkey\n";
 	print STDERR "$name Playing back recorded file.\n";
 	print "STREAM FILE $tmpname \"\"\n";
-	@result = &checkresponse();
+	@result = checkresponse();
 	print STDERR "$name Failed to play file\n" if ($result[0] == -1);
 }
 
@@ -279,7 +279,7 @@ print STDERR "$name The response was:\n", $uaresponse->content if ($debug);
 foreach (keys %response) {
 	print STDERR "$name Setting variable: $_ = $response{$_}\n" if ($debug);
 	print "SET VARIABLE \"$_\" \"$response{$_}\"\n";
-	&checkresponse();
+	checkresponse();
 }
 exit;