Audio Recognition Python SDK

Overview

ACRCloud provides services such as Music Recognition, Broadcast Monitoring, Custom Audio Recognition, Copyright Compliance & Data Deduplication, Live Channel Detection, and Offline Recognition etc.

This audio recognition python SDK support most of audio / video files.

Audio: mp3, wav, m4a, flac, aac, amr, ape, ogg ...
Video: mp4, mkv, wmv, flv, ts, avi ...

Requirements

Follow one of the tutorials to create a project and get your host, access_key and access_secret.

• Recognize Music
• Recognize Custom Content
• Broadcast Monitoring for Music
• Broadcast Monitoring for Custom Content
• Detect Live & Timeshift TV Channels
• Recognize Custom Content Offline
• Recognize Live Channels and Custom Content

Install

Linux and macOS

python3 -m pip install pyacrcloud

Other platform

You can run python -m pip install git+https://github.com/acrcloud/acrcloud_sdk_python or go to sub dir, and run"sudo python setup.py install".

Windows Runtime Library

If you run the SDK on Windows, you must install this library.
X86: download and install Library(windows/vcredist_x86.exe)
x64: download and install Library(windows/vcredist_x64.exe)

Note

1. If you run the SDK on Windows, you must install library(vcredist).
2. ALL version supports humming.
3. If you use docker alpine, you need to install "apk add --update libstdc++"

Functions

Introduction all API.

recognizer.py

classACRCloudRecognizer:
defrecognize_by_file(self, file_path, start_seconds, rec_length=10):
#@param file_path : query file path#@param start_seconds : skip (start_seconds) seconds from from the beginning of (filePath)#@param rec_length: use rec_length seconds data to recongize#@return result metainfosdefrecognize_by_filebuffer(self, file_buffer, start_seconds, rec_length=10):
#@param file_buffer : file_path query buffer#@param start_seconds : skip (start_seconds) seconds from from the beginning of (filePath)#@param rec_length: use rec_length seconds data to recongize#@return result metainfosdefrecognize(self, wav_audio_buffer):
#@param wav_audio_buffer : query buffer(RIFF (little-endian) data, WAVE audio, Microsoft PCM, 16 bit, mono 8000 Hz)#@return result metainfos

Module acrcloud_extr_tool

defcreate_fingerprint_by_file(file_name, start_time_seconds, audio_len_seconds, is_db_fingerprint, opt):
#file_name: Path of input file; #start_time_seconds: Start time of input file, default is 0; #audio_len_seconds: Length of audio data you need. if you create recogize frigerprint, default is 12 seconds, if you create db frigerprint, it is not usefully; #is_db_fingerprint: If it is True, it will create db frigerprint (Fingerprint for bucket, not for recognition); #opt opt = {'filter_energy_min': 0,
'silence_energy_threshold': 100,
'silence_rate_threshold': 1
}
defcreate_humming_fingerprint_by_file(file_name, start_time_seconds, audio_len_seconds):
#file_name: Path of input file; #start_time_seconds: Start time of input file, default is 0; #audio_len_seconds: Length of audio data you need. if you create recogize frigerprint, default is 12 seconds, if you create db frigerprint, it is not usefully; defcreate_fingerprint_by_filebuffer(data_buffer, start_time_seconds, audio_len_seconds, is_db_fingerprint, opt):
#data_buffer: data buffer of input file; #start_time_seconds: Start time of input file, default is 0; #audio_len_seconds: Length of audio data you need. if you create recogize frigerprint, default is 12 seconds, if you create db frigerprint, it is not usefully; #is_db_fingerprint: If it is True, it will create db frigerprint (Fingerprint for bucket, not for recognition); #opt = {'filter_energy_min': 0,
'silence_energy_threshold': 100,
'silence_rate_threshold': 1
}
defcreate_humming_fingerprint_by_filebuffer(data_buffer, start_time_seconds, audio_len_seconds):
#data_buffer: data buffer of input file; #start_time_seconds: Start time of input file, default is 0; #audio_len_seconds: Length of audio data you need. if you create recogize frigerprint, default is 12 seconds, if you create db frigerprint, it is not usefully; defcreate_fingerprint(data_buffer, is_db_fingerprint):
#data_buffer: audio data buffer(RIFF (little-endian) data, WAVE audio, Microsoft PCM, 16 bit, mono 8000 Hz); #is_db_fingerprint: If it is True, it will create db frigerprint (Fingerprint for bucket, not for recognition); defcreate_humming_fingerprint(data_buffer):
#data_buffer: audio data buffer(RIFF (little-endian) data, WAVE audio, Microsoft PCM, 16 bit, mono 8000 Hz); defdecode_audio_by_file(file_name, start_time_seconds, audio_len_seconds):
#It will return the audio data(RIFF (little-endian) data, WAVE audio, Microsoft PCM, 16 bit, mono 8000 Hz); #file_name: Path of input file; #start_time_seconds: Start time of input file, default is 0; #audio_len_seconds: Length of audio data you need, if it is 0, will decode all the audio; defdecode_audio_by_filebuffer(data_buffer, start_time_seconds, audio_len_seconds):
#It will return the audio data(RIFF (little-endian) data, WAVE audio, Microsoft PCM, 16 bit, mono 8000 Hz); #data_buffer: data buffer of input file; #start_time_seconds: Start time of input file, default is 0; #audio_len_seconds: Length of audio data you need, if it is 0, will decode all the audio; defversion() #return the version of this module

Example

run Test: python test.py test.mp3

#!/usr/bin/env python#-*- coding:utf-8 -*-importos, sysfromacrcloud.recognizerimportACRCloudRecognizerif__name__=='__main__':
config= {
#Replace "xxxxxxxx" below with your project's host, access_key and access_secret.'host':'XXXXXXXX',
'access_key':'XXXXXXXX', 'access_secret':'XXXXXXXX',
'timeout':10# seconds
}
'''This module can recognize ACRCloud by most of audio/video file.  Audio: mp3, wav, m4a, flac, aac, amr, ape, ogg ... Video: mp4, mkv, wmv, flv, ts, avi ...'''re=ACRCloudRecognizer(config)
# process each for 10 seconds.filepath=sys.argv[1]
duration_ms=int(re.get_duration_ms_by_file(filepath))
foriinrange(0, duration_ms//1000, 10):
res=re.recognize_by_file(filepath, i, 10)
print(i, res.strip())
#recognize by file path, and skip 0 seconds from from the beginning of sys.argv[1].printre.recognize_by_file(sys.argv[1], 0)
buf=open(sys.argv[1], 'rb').read()
#recognize by file_audio_buffer that read from file path, and skip 0 seconds from from the beginning of sys.argv[1].printre.recognize_by_filebuffer(buf, 0)