# This is part of Kaylee # -- this code is licensed GPLv3 # Copyright 2013 Jezra # Copyright 2015 Clayton G. Hobbs import gi gi.require_version('Gst', '1.0') from gi.repository import GObject, Gst GObject.threads_init() Gst.init(None) import os.path import sys # Define some global variables this_dir = os.path.dirname( os.path.abspath(__file__) ) class Recognizer(GObject.GObject): __gsignals__ = { 'finished' : (GObject.SIGNAL_RUN_LAST, GObject.TYPE_NONE, (GObject.TYPE_STRING,)) } def __init__(self, language_file, dictionary_file, src = None): GObject.GObject.__init__(self) self.commands = {} if src: audio_src = 'alsasrc device="hw:%d,0"' % (src) else: audio_src = 'autoaudiosrc' # Build the pipeline cmd = audio_src+' ! audioconvert ! audioresample ! pocketsphinx name=asr ! appsink sync=false' try: self.pipeline=Gst.parse_launch( cmd ) except Exception, e: print e.message print "You may need to install gstreamer1.0-pocketsphinx" raise e bus = self.pipeline.get_bus() bus.add_signal_watch() # Get the Auto Speech Recognition piece asr=self.pipeline.get_by_name('asr') bus.connect('message::element', self.result) asr.set_property('lm', language_file) asr.set_property('dict', dictionary_file) asr.set_property('configured', True) def listen(self): self.pipeline.set_state(Gst.State.PLAYING) def pause(self): self.pipeline.set_state(Gst.State.PAUSED) def result(self, bus, msg): msg_struct = msg.get_structure() # Ignore messages that aren't from pocketsphinx msgtype = msg_struct.get_name() if msgtype != 'pocketsphinx': return # If we have a final command, send it for processing command = msg_struct.get_string('hypothesis') if command != '' and msg_struct.get_boolean('final')[1]: self.emit("finished", command)