Merge pull request #49 from verbio-technologies/feature/SC2_1216_add_…

…alaw_support

Added support for a-law audios to be able to use audios from Barrayar.

cli-client/helpers/audio_importer.py

@@ -1,11 +1,27 @@
+import os
+import subprocess
 import wave
 
+def preprocess_audio_file_to_pcm(audio_file: str):
+    tmp_audio_file = "./" + os.path.basename(audio_file) + "_tmp.wav"
+    command = "sox " + audio_file + " -e signed-integer " + tmp_audio_file
+    subprocess.run(command.split())
+    return tmp_audio_file
+
+
+def remove_pcm_audio_file(audio_file: str):
+    os.remove(audio_file)
+
 
 class AudioImporter:
-    def __init__(self, audio_file: str):
-        with open(audio_file, "rb") as wav_file:
+    def __init__(self, audio_file: str, convert_audio: bool):
+        tmp_audio_file = audio_file
+        if convert_audio:
+            tmp_audio_file = preprocess_audio_file_to_pcm(audio_file)
+        with open(tmp_audio_file, "rb") as wav_file:
             wav_data = wave.open(wav_file)
             self.sample_rate = wav_data.getframerate()
             self.audio = wav_data.readframes(wav_data.getnframes())
             wav_data.close()
-
+        if convert_audio:
+            remove_pcm_audio_file(tmp_audio_file)

cli-client/helpers/common.py

@@ -125,6 +125,8 @@ def parse_csr_commandline() -> RecognizerOptions:
     options = RecognizerOptions()
     parser = argparse.ArgumentParser(description='Perform speech recognition on an audio file')
     parser.add_argument('--audio-file', '-a', help='Path to a .wav audio in 8kHz and PCM16 encoding', required=True)
+    parser.add_argument('--convert-audio', '-c', help='Convert audio file to from A-LAW to PCM using sox software. Used for internal testing.',
+                        required=False, default=False, dest='convert_audio', action='store_true')
     topicGroup = parser.add_mutually_exclusive_group(required=True)
     topicGroup.add_argument('--topic', '-T', choices=['GENERIC', 'TELCO', 'BANKING', 'INSURANCE'], help='A valid topic')
     topicGroup.add_argument('--inline-grammar', '-I', help='Grammar inline as a string')
@@ -172,6 +174,7 @@ def parse_csr_commandline() -> RecognizerOptions:
     options.token_file = args.token
     options.host = args.host
     options.audio_file = args.audio_file
+    options.convert_audio = args.convert_audio
     options.language = args.language
     options.secure_channel = args.secure
     options.formatting = args.formatting

cli-client/recognizer_stream.py

@@ -12,7 +12,7 @@
 
 
 def process_recognition(executor: ThreadPoolExecutor, channel: grpc.Channel, options: RecognizerOptions, access_token: str):
-    audio_resource = AudioImporter(options.audio_file)
+    audio_resource = AudioImporter(options.audio_file, options.convert_audio)
     stub = recognition_pb2_grpc.RecognizerStub(channel)
     client = CSRClient(executor, stub, options, audio_resource, access_token)
     client.send_audio()