Purpose

Listen to television audio, detect commercials, request decrease volume.

Results

install Dejavu

conda activate remy_python
pip install -e git+https://github.com/bcollazo/dejavu@v1.2#egg=PyDejavu

This installed directory src/pydejavu. I didn't delete file added by pip: src/pip-delete-this-directory.txt See also Appendix- Install dejavu dependencies.

Activate conda environment

If using Anaconda.

conda activate <my_environment>
conda activate remy_python

Install audio files

convert wav to mp3

ffmpeg -i a.wav -f mp3 a.mp3

put audio mp3 files in directory data/commercials_mp3

Select live audio input source

Selecting live audio input from an audio interface instead of from microphone gives a cleaner signal without room noise. If using macOS and an audio interface, go to System Preferences / Sound / Input. Select the interface e.g. Scarlett 2i4 USB. Note I haven't tried running Dejavu on Raspberry Pi Raspbian yet, and haven't selected an audio input source.

run dejavu test script

Optional step. Running test_dejavu.sh populates results directory including test.db Ok to delete contents of results directory.

$ chmod u+x ./src/pydejavu/test_dejavu.sh
$ ./src/pydejavu/test_dejavu.sh

database

check sqlite schema

database url format

https://stackoverflow.com/questions/372885/how-do-i-connect-to-a-mysql-database-in-python?rq=1

mysql://<user_name>:<password>@<host_name>/<db_name>
mysql+mysqlconnector://beepscore:password@localhost/dejavu

set environment variable for database

data/config.json

may contain a json dictionary e.g.

{"DATABASE_URL": "results/test.db"}

Ways to run: Can run recognize_audio.py If no database exists it will make a new one, I think in-memory sqlite.

conda activate <environment>
python3 recognize_audio.py

example output

2019-04-21 18:42:17 DEBUG    recognize_audio_from_a_file line:49
filename_containing_audio_to_match: data/commercial_mp3/chantix.mp3,
match_dict_json: {"song_id": 14, "song_name": "chantix", "confidence": 43335,
"offset": 0, "offset_seconds": 0.0, "file_sha1": "a4fcabb2afb6518a65321c5807c5fc7c798a0aa2",
"match_time": 11.153626918792725}

2019-04-21 18:42:48 DEBUG    recognize_audio_from_microphone line:80
From mic with 5 seconds we recognized: {"song_id": 19,
"song_name": "tecovas-western-boots", "confidence": 179, "offset": 62,
"offset_seconds": 2.87927, "file_sha1": "ca324b141675b59c05bdcbbdf05c038d68940376"}

mysql database

I used conda to install mysql-connector-python https://stackoverflow.com/questions/34168651/what-are-the-differences-between-mysql-connector-python-mysql-connector-python

Then ran recognize_audio.py

Got error "Can't connect to MySQL server on 'localhost' (61)" To use mysql, may need to start mysql server before running recognize_audio.py https://stackoverflow.com/questions/119008/cant-connect-to-mysql-server-on-localhost-10061

unit tests

Note: unittest discover requires tests directory contain init.py cd to project root directory

conda activate remy_python
python -m unittest discover

Appendix- Install dejavu dependencies

I installed many of these as conda packages in environment remy_python. I don't know if installing dejavu from pip would handle this automatically. https://github.com/worldveil/dejavu/blob/master/INSTALLATION.md

ffmpeg

for converting audio files to .wav format on macOS

brew install ffmpeg

matplotlib

used for spectrograms and plotting

MySQLdb

for interfacing with MySQL databases

numpy

for taking the FFT of audio signals

pydub

a Python ffmpeg wrapper

scipy

used in peak finding algorithms

pyaudio

for grabbing audio from microphone

sqlalchemy

sqlalchemy_utils

https://sqlalchemy-utils.readthedocs.io/en/latest/installation.html

package name contains dash

sqlalchemy-utils

import name contains underscore

import sqlalchemy_utils

References

Python Data Analysis to Automatically Detect and Mute Television Commercials

http://beepscore.com/website/2019/04/21/automatically-detecting-television-commercials.html

Dejavu project

Audio fingerprinting and recognition in Python. It has recognizers for recorded files and for live microphone input. Originally by worldveil/dejavu, forked by DataWookie/dejavu and bcollazo/dejavu. In 2019-03 bcollazo/dejavu appeared to be the most active fork.

beepscore/Dejavu (currently unused)

I forked bcollazo/dejavu as beepscore/dejavu and added branch "beepscore". https://github.com/beepscore/dejavu/tree/dejavu

PyAudio with USB audio interface

Recording multiple microphones in python

Scarlett audio interface and PyAudio https://stackoverflow.com/questions/25620285/recording-multiple-microphones-in-python

python2.7 on Raspberry Pi 3 - Pyaudio Input overflowed

Raspberry Pi 3 Model B and Focusrite Scarlett 2i2 USB Sound Card https://stackoverflow.com/questions/38042875/python2-7-on-raspberry-pi-3-pyaudio-input-overflowed

remy_python

A Raspberry Pi infrared remote control. The Python app has three parts: Functions to send commands to the infrared transmitter. A Flask web service to accept television command requests (e.g. volume decrease, volume increase). A scheduler that automatically sends remote control commands at programmed times (e.g. mute during TV commercials). https://github.com/beepscore/remy_python

Network enabled Raspberry Pi tv remote control

http://beepscore.com/website/2019/02/05/network-enabled-raspberry-pi-tv-remote-control.html

Remy

Remote control television by sending commands from iOS device to a server. https://github.com/beepscore/Remy

Appendix- update Dejavu to use logging_util compatible with tv_commercial_silencer

I used pip to install src/pydejavu and it isn't committed to this git repo. Manually edited src/pydejavu/dejavu/init.py

# import logging
import logging_util
...

# logger = logging.getLogger(__name__)
logger = logging_util.get_logger(__name__)