Python Alexa Voice Service App

Features

Support Alexa Voice Service API v20160207
Support multiple audio players: gstreamer 1.0, mpv and mpg123
支持Baidu DuerOS

Options

Player

We have 3 players (mpv, mpg123 and gstreamer) to use. SpeechSynthesizer and Alerts prefer mpg123 which is more responsive. AudioPlayer likes gstreamer > mpv > mpg123. Gstreamer supports more audio format and works well on raspberry pi. We can also specify the player of AudioPlayer using the environment variable PLAYER.
Recorder

2 recorders (pyaudio & arecord) are available. We can use environment variable RECORDER to specify the recorder. For example, run RECORDER=pyaudio alexa-tap will use pyaudio as the recorder. By default, arecord is used as the recorder.
Keyword detector (optional)

Use PocketSphinx or Snowboy. To use pocketsphinx, install respeaker python library and pocketsphinx. To use Snowboy, go to Snowboy's Github to install it.

If you use raspberry pi and gstreamer, it is likely that gstreamer's default audio sink is GstOMXHdmiAudioSink. It ignores ALSA configurations and outputs audio to HDMI. If you don't want to use HDMI audio output, you should run sudo apt remove gstreamer1.0-omx gstreamer1.0-omx-rpi

Requirements

For ReSpeaker Core (MT7688)

gstreamer1.0, pyaudio and pocketsphinx and respeaker python library are already installed by default, just run pip install avs

For Debian/Ubuntu/Raspbian

sudo apt-get install mpg123 mpv
sudo apt-get install gstreamer1.0-plugins-good gstreamer1.0-plugins-bad gstreamer1.0-plugins-ugly \
gir1.2-gstreamer-1.0 python-gi python-gst-1.0
sudo apt-get install python-pyaudio

Get started

run alexa-audio-check to check if recording & playing is OK. If RMS is not zero, recording is OK, if you can hear alarm, playing is OK
```
 $alexa-audio-check
 RMS: 41
 RMS: 43
```
run alexa-auth to login Amazon, it will save authorization information to ~/.avs.json
run alexa-tap, then press Enter to talk with alexa

If you want to use a specified player, use the environment variable PLAYER to specify it, such as PLAYER=mpv alexa-tap or PLAYER=mpg123 alexa or PLAYER=gstreamer alexa

Hands-free Alexa

Using PocketSphinx for Keyword Spotting

install respeaker and pocketsphinx python packages

sudo pip install respeaker pocketsphinx # pocketsphinx requires gcc toolchain and libpulse-dev
run alexa, then use "alexa" to start a conversation with alexa, for example, "alexa, what time is it"

Using Snowboy for Keyword Spotting

Install Snowboy

git clone --depth 1 https://github.com/Kitt-AI/snowboy.git snowboy_github
cd snowboy_github
sudo apt install libatlas-base-dev swig
python setup.py build
sudo pip install .

Install voice-engine python library

sudo pip install voice-engine

run the following python script and use the keyword alexa to start a conversation with alexa

import time
import signal
from voice_engine.source import Source
from voice_engine.kws import KWS
from avs.alexa import Alexa
import logging

logging.basicConfig(level=logging.DEBUG)


src = Source(rate=16000)
kws = KWS(model='snowboy')
alexa = Alexa()

src.pipeline(kws, alexa)

def on_detected(keyword):
    print('detected {}'.format(keyword))
    alexa.listen()

kws.set_callback(on_detected)

is_quit = []
def signal_handler(signal, frame):
    print('Quit')
    is_quit.append(True)

signal.signal(signal.SIGINT, signal_handler)

src.pipeline_start()
while not is_quit:
    time.sleep(1)
src.pipeline_stop()

To do

Speaker interface
Notifications interface

Change Alexa Voice Service client id and product id

If you want to use your own client id and product id, try:

create a file named config.json with your product_id, client_id and client_secret

 {
     "product_id": "x",
     "client_id": "y",
     "client_secret": "z"
 }

run alexa-auth -c config.json
run alexa-tap or alexa

License

GNU General Public License v3

Credits

This project is based on nicholas-gh/python-alexa-client

This package was created with Cookiecutter_ and the audreyr/cookiecutter-pypackage project template.

respeaker/avs