added timeout on speech recognition

CMakeLists.txt

@@ -2,6 +2,8 @@ cmake_minimum_required(VERSION 2.8.3)
 project(respeaker_ros)
 
 find_package(catkin REQUIRED COMPONENTS
+  rospy
+  message_generation
   catkin_virtualenv
   dynamic_reconfigure)
 if ($ENV{ROS_PYTHON_VERSION} EQUAL 2)
@@ -12,10 +14,21 @@ else()
     "roscd respeaker_ros && pip install -r requirements.txt")
 endif()
 
+add_message_files(
+  DIRECTORY
+  msg
+  FILES
+  StampedAudio.msg
+)
+
+generate_messages()
+
 generate_dynamic_reconfigure_options(
   cfg/Respeaker.cfg)
 
-catkin_package()
+catkin_package(
+  CATKIN_DEPENDS message_runtime
+)
 
 if (catkin_virtualenv_FOUND)
   catkin_generate_virtualenv()
@@ -37,4 +50,4 @@ install(FILES requirements.txt
 if (CATKIN_ENABLE_TESTING)
   catkin_add_nosetests(test/test_installed.py
     DEPENDENCIES ${PROJECT_NAME}_generate_virtualenv)
-endif()
+endif()

package.xml

@@ -3,11 +3,13 @@
   <name>respeaker_ros</name>
   <version>0.0.0</version>
   <description>The respeaker_ros package</description>
-  <maintainer email="[email protected]">Yuki Furuta</maintainer>
+  <maintainer email="[email protected]">Zachary Kratochvil</maintainer>
   <license>Apache</license>
   <author email="[email protected]">Yuki Furuta</author>
 
   <buildtool_depend>catkin</buildtool_depend>
+  <build_depend>message_generation</build_depend>
+
   <depend condition="$ROS_PYTHON_VERSION == 3">catkin_virtualenv</depend>
   <depend>angles</depend>
   <depend>audio_common_msgs</depend>
@@ -20,7 +22,11 @@
   <depend>speech_recognition_msgs</depend>
   <depend>std_msgs</depend>
   <depend>tf</depend>
+
   <export>
     <pip_requirements>requirements.txt</pip_requirements>
   </export>
+
+  <exec_depend>message_runtime</exec_depend>
+
 </package>

scripts/respeaker_node.py

@@ -17,7 +17,7 @@
 import sys
 import time
 import speech_recognition as SR
-from audio_common_msgs.msg import AudioData
+from respeaker_ros.msg import StampedAudio
 from geometry_msgs.msg import PoseStamped
 from std_msgs.msg import Bool, Int32, ColorRGBA
 from dynamic_reconfigure.server import Server
@@ -340,9 +340,9 @@ def __init__(self):
         self.pub_vad = rospy.Publisher("is_speeching", Bool, queue_size=1, latch=True)
         self.pub_doa_raw = rospy.Publisher("sound_direction", Int32, queue_size=1, latch=True)
         self.pub_doa = rospy.Publisher("sound_localization", PoseStamped, queue_size=1, latch=True)
-        self.pub_audio = rospy.Publisher("audio", AudioData, queue_size=10)
-        self.pub_speech_audio = rospy.Publisher("speech_audio", AudioData, queue_size=10)
-        self.pub_audios = {c:rospy.Publisher('audio/channel%d' % c, AudioData, queue_size=10) for c in self.respeaker_audio.channels}
+        self.pub_audio = rospy.Publisher("audio", StampedAudio, queue_size=10)
+        self.pub_speech_audio = rospy.Publisher("speech_audio", StampedAudio, queue_size=2)
+        self.pub_audios = {c:rospy.Publisher('audio/channel%d' % c, StampedAudio, queue_size=10) for c in self.respeaker_audio.channels}
         # init config
         self.config = None
         self.dyn_srv = Server(RespeakerConfig, self.on_config)
@@ -401,9 +401,9 @@ def on_audio(self, data, channel):
         if channel == 0:
             self.out.writeframes(data)
 
-        self.pub_audios[channel].publish(AudioData(data=data))
+        self.pub_audios[channel].publish(StampedAudio(data=data, stamp=rospy.get_rostime()))
         if channel == self.main_channel:
-            self.pub_audio.publish(AudioData(data=data))
+            self.pub_audio.publish(StampedAudio(data=data, stamp=rospy.get_rostime()))
             if self.is_speeching:
                 if len(self.speech_audio_buffer) == 0:
                     self.speech_audio_buffer = self.speech_prefetch_buffer
@@ -456,7 +456,7 @@ def on_timer(self, event):
             rospy.loginfo("Speech detected for %.3f seconds" % duration)
             if self.speech_min_duration <= duration < self.speech_max_duration:
 
-                self.pub_speech_audio.publish(AudioData(data=list(buf)))
+                self.pub_speech_audio.publish(StampedAudio(data=list(buf), stamp=rospy.get_rostime()))
 
 
 if __name__ == '__main__':

scripts/speech_to_text.py

@@ -7,7 +7,7 @@
 import speech_recognition as SR
 
 from actionlib_msgs.msg import GoalStatus, GoalStatusArray
-from audio_common_msgs.msg import AudioData
+from respeaker_ros.msg import StampedAudio
 from sound_play.msg import SoundRequest, SoundRequestAction, SoundRequestGoal
 from speech_recognition_msgs.msg import SpeechRecognitionCandidates
 
@@ -40,8 +40,8 @@ def __init__(self):
                 self.tts_action = None
 
         self.pub_speech = rospy.Publisher(
-            "speech_to_text", SpeechRecognitionCandidates, queue_size=1)
-        self.sub_audio = rospy.Subscriber("speech_audio", AudioData, self.audio_cb)
+            "speech_to_text", SpeechRecognitionCandidates, queue_size=2)
+        self.sub_audio = rospy.Subscriber("speech_audio", StampedAudio, self.audio_cb, queue_size=2)
 
     def tts_timer_cb(self, event):
         stamp = event.current_real
@@ -63,9 +63,14 @@ def tts_timer_cb(self, event):
                 self.is_canceling = Falser
 
     def audio_cb(self, msg):
+
         if self.is_canceling:
             rospy.loginfo("Speech is cancelled")
             return
+        if rospy.get_rostime().secs - msg.stamp.secs > 2:
+            rospy.loginfo("Old speech discarded")
+            return
+
         data = SR.AudioData(bytes(msg.data), self.sample_rate, self.sample_width)
         with open(str(len(msg.data)) + ".wav","wb") as f:
             f.write(data.get_wav_data())