Merge pull request #471 from thomasgubler/espeak

linux speech synthesis: switch to eSpeak

QGCExternalLibs.pri

@@ -523,7 +523,7 @@ MacBuild {
 ##
 # [OPTIONAL] Speech synthesis library support.
 # Can be forcibly disabled by adding a `DEFINES+=DISABLE_SPEECH` argument to qmake.
-# Linux support requires the Festival Lite speech synthesis engine (flite).
+# Linux support requires the eSpeak speech synthesizer (espeak).
 # Mac support is provided in Snow Leopard and newer (10.6+)
 # Windows is supported as of Windows 7
 #
@@ -534,18 +534,11 @@ contains (DEFINES, DISABLE_SPEECH) {
 } else:infile(user_config.pri, DEFINES, DISABLE_SPEECH) {
     message("Skipping support for speech output (manual override from user_config.pri)")
 } else:LinuxBuild {
-	exists(/usr/include/flite) | exists(/usr/local/include/flite) {
+	exists(/usr/include/espeak) | exists(/usr/local/include/espeak) {
 		message("Including support for speech output")
 		DEFINES += QGC_SPEECH_ENABLED
 		LIBS += \
-		-lflite_cmu_us_kal \
-		-lflite_usenglish \
-		-lflite_cmulex \
-		-lflite
-
-		# We need to add the alsa asound library as well for some Linux platforms
-		# (like Arch)
-		LIBS += -lasound
+		-lespeak
 	} else {
 		warning("Skipping support for speech output (missing libraries, see README)")
 	}

README.md

@@ -44,7 +44,7 @@ An add-on is available for QGC that provides a UI for generating MAVLink dialect
 Integration with Opal-RT's RT-LAB simulator can be enabled on Windows by installing RT-LAB 7.2.4. This allows vehicles to be simulated in RT-LAB and communicate directly with QGC on the same computer as if the UAS was actually deployed. This support is enabled by default once the requisite RT-LAB software is installed. Disabling this can be done by adding `DISABLE_RTLAB` to the `DEFINES` variable.
 
 ### Speech syntehsis
-QGroundcontrol can notify the controller of information via speech synthesis. This requires the `flite` library on Linux. On Mac and Windows support is built in to the OS as of OS X 10.6 (Snow Leopard) and Windows Vista. This support is enabled by default on all platforms if the dependencies are met. Disabling this functionality can be done by adding `DISABLE_SPEECH` to the `DEFINES` variable.
+QGroundcontrol can notify the controller of information via speech synthesis. This requires the `espeak` library on Linux. On Mac and Windows support is built in to the OS as of OS X 10.6 (Snow Leopard) and Windows Vista. This support is enabled by default on all platforms if the dependencies are met. Disabling this functionality can be done by adding `DISABLE_SPEECH` to the `DEFINES` variable.
 
 ### 3D view
 The OpenSceneGraph libraries provide 3D rendering to the map overlays that QGC can provide.
@@ -92,9 +92,10 @@ To build on Linux:
   * For Fedora: `sudo yum install qt qt-creator qt-webkit-devel SDL-devel SDL-static systemd-devel`
 
 2. **[OPTIONAL]** Install additional libraries
-  * For text-to-speech (flite)
-	* For Ubuntu: `sudo apt-get install libflite1 flite1-dev`
-	* For Fedora: `sudo yum install flite-devel`
+  * For text-to-speech (espeak)
+	* For Ubuntu: `sudo apt-get install espeak libespeak-dev`
+	* For Fedora: `sudo yum install espeak espeak-devel`
+	* For Arch Linux: `pacman -Sy espeak`
   * For 3D flight view (openscenegraph)
 	* For Ubuntu: `sudo apt-get install libopenscenegraph-dev`
 	* For Fedora: `sudo yum install OpenSceneGraph-qt-devel`

src/GAudioOutput.cc

@@ -26,6 +26,7 @@ This file is part of the QGROUNDCONTROL project
  *   @brief Implementation of audio output
  *
  *   @author Lorenz Meier <mavteam@student.ethz.ch>
+ *   @author Thomas Gubler <thomasgubler@gmail.com>
  *
  */
 
@@ -51,10 +52,8 @@ This file is part of the QGROUNDCONTROL project
 #endif
 
 #if defined Q_OS_LINUX && defined QGC_SPEECH_ENABLED
-extern "C" {
-#include <flite/flite.h>
-    cst_voice *register_cmu_us_kal(const char *voxdir);
-};
+// Using eSpeak for speech synthesis: following https://github.com/mondhs/espeak-sample/blob/master/sampleSpeak.cpp
+#include <espeak/speak_lib.h>
 #endif
 
 #if defined _MSC_VER && defined QGC_SPEECH_ENABLED
@@ -97,7 +96,12 @@ GAudioOutput::GAudioOutput(QObject *parent) : QObject(parent),
 
 
 #if defined Q_OS_LINUX && defined QGC_SPEECH_ENABLED
-    flite_init();
+    espeak_Initialize(AUDIO_OUTPUT_PLAYBACK, 500, NULL, 0); // initialize for playback with 500ms buffer and no options (see speak_lib.h)
+    espeak_VOICE espeak_voice = {};
+    espeak_voice.languages = "en-uk"; // Default to British English
+    espeak_voice.name = "klatt"; // espeak voice name
+    espeak_voice.gender = 2; // Female
+    espeak_SetVoiceByProperties(&espeak_voice);
 #endif
 
 #if defined _MSC_VER && defined QGC_SPEECH_ENABLED
@@ -206,20 +210,8 @@ bool GAudioOutput::say(QString text, int severity)
 #endif
 
 #if defined Q_OS_LINUX && defined QGC_SPEECH_ENABLED
-            QTemporaryFile file;
-            file.setFileTemplate("XXXXXX.wav");
-
-            if (file.open())
-            {
-                cst_voice *v = register_cmu_us_kal(NULL);
-                cst_wave *wav = flite_text_to_wave(text.toStdString().c_str(), v);
-                // file.fileName() returns the unique file name
-                cst_wave_save(wav, file.fileName().toStdString().c_str(), "riff");
-                m_media->setCurrentSource(Phonon::MediaSource(QUrl::fromLocalFile(file.fileName().toStdString().c_str())));
-                m_media->play();
-                res = true;
-            }
-
+            unsigned int espeak_size = strlen(text.toStdString().c_str());
+            espeak_Synth(text.toStdString().c_str(), espeak_size, 0, POS_CHARACTER, 0, espeakCHARS_AUTO, NULL, NULL);
 #endif
 
 #if defined Q_OS_MAC && defined QGC_SPEECH_ENABLED

src/GAudioOutput.h

@@ -40,9 +40,6 @@ This file is part of the PIXHAWK project
 #include <AudioOutput>
 #endif
 #ifdef Q_OS_LINUX
-#if defined QGC_SPEECH_ENABLED
-//#include <flite/flite.h>
-#endif
 #include <phonon/MediaObject>
 #include <phonon/AudioOutput>
 #endif