fat84 / audioRecognition

Audioneex-OS is a general purpose audio content recognition engine designed for real-time uses.

Audioneex-OS is an audio content recognition engine specifically designed for real-time applications. It is general purpose, based on content-agnostic algorithms and runs on desktop, server, mobile and embedded devices.

Features

• Compact fingerprints - 1 hr of audio encoded in less than a MB
• Content-agnostic - recognition of audio of different nature
• Fast identification - can be used for real-time recognitions
• Cross-platform - runs anywhere there is a fully compliant C++11 compiler
• IoT & Mobile-ready - runs well on small devices for on-device ACR
• Database-agnostic - can use different databases by rewriting the drivers

Limitations

• Limited scaling on a single instance
• Limited robustness to noisy over-the-air audio
• Documentation is (currently) not available

Prerequisites

The engine needs, and has been tested with, the following dependencies

• Boost 1.6
• FFTSS 3.0
• Tokyo Cabinet 1.4 / Couchbase 5.1
• TagLib (optional)
• FFMpeg (optional)

and the following tools

• GCC 6/7, CLang 5, MSVC++ 14
• CMake 3.11
• Android NDK r16b

However, it should also work with any other fully compliant C++11 compiler. The TagLib and FFMpeg can be replaced with something else but you will need to make changes to the code. Support for other databases other than the default ones (Tokyo Cabinet/Couchbase) can be added by reimplementing the exposed interfaces and following the specifications for the drivers.

How to build on Linux and Windows

The project uses the CMake build system on both Linux and Windows. The steps to follow are pretty much the same on both platforms, aside for few specific tweaks that may occur.

Step 1. Set up the build environment

• Install Boost. The library requires the header-only part, but the examples will need some compiled modules (thread, filesystem, regex and their dependencies).

-fPIC flag on Linux otherwise linking errors will occur).

• Get the Tokyo Cabinet library (for Windows there is a port from the EJDB project). Alternatively, you can use the Couchbase database (requires the libcouchbase driver).
• Get the TagLib library for ID3 tag support (used by the examples to extract metadata from audio files).
• Get the FFmpeg executable and install it in a location visible from PATH (used by the examples to decode and read audio).

Step 2. Set include and library paths

This step is not mandatory but it will most likely be necessary since these paths are system-dependent. You can set them in the "User Configuration" section of the CMake build script located in the root directory.

Step 3. Build

On Linux, open a shell and issue the following commands

cd <source_root_directory>
mkdir build && cd build
cmake [options] ..
make

where [options] is one or more of the following command line parameters in the form -D<var=value>

ARCH            = x32|x64
TOOLCHAIN       = gcc64|vc14|...
BINARY_TYPE     = dynamic|static
BUILD_MODE      = debug|release
WITH_EXAMPLES   = ON|OFF
DATASTORE_T     = TCDataStore|CBDataStore  (for the examples only)
ID3_TAG_SUPPORT = ON|OFF  (for the examples only)
WITH_TESTS      = ON|OFF  (for project developers only)

On Windows, replace the last two commands above with the following

cmake -G "NMake Makefiles" [options] ..
nmake

By default, if no options are passed on, the script builds dynamic libraries targeting 64-bit architectures release mode. The final libraries will be put in a /lib folder in the root directory.

How to build for Android

There is a build script in the root directory called build_android that will ease the compilation of the library for Android platforms. This script uses the Android NDK build system (ndk-build et al.) to build the library without the need for exporting individual toolchains for each architecture.

Usage:

build_android <arch> <comp> <api> <bmode> <btype>

where

<arch>   is one of the supported architectures (armeabi-v7a, x86, etc.)
<comp>   the compiler (clang, gcc)
<api>    the target Android API version
<bmode>  the build mode (debug, release)
<btype>  the library type (static, dynamic)

The final libraries will be put in the /lib folder of the root directory.

As a prerequisite, you will have to build the required external libraries for the specific Android platforms you're targeting. There is a build script in the root directory called android-configure that will help you with the cross-compilation of the libraries without the need for exporting toolchains for each target architecture. For more info, have a look at the script itself.

This script has been tested with the NDK r16b. Please refer to the script for more information (especially for how to fix some bugs present in r16b).

License

This code is released under the Mozilla Public License 2.0.

In a nutshell:

• You can use it in commercial projects without publishing your closed source code
• If you distribute it publicly under any form (code or binary, stand-alone or combined) you must make its source code available, including any modification you have made, under the same license, retain all copyright notices and inform the users where they can get it.

TODO

1. More drivers for other databases would be nice
2. A more comprehensive test suite
3. Use more C++ new features and drop old compilers support