Open Binaural Renderer (obr)

The obr binaural rendering library is designed to render immersive audio content.

Features

The library is written using C++ and it utilizes Bazel build system. A CMake configuration is also provided.

The library renders the following types of Audio Elements:

Ambisonics (from 1st to 4th order).
Channel-based (e.g. 5.1, 7.1.4). See Loudspeaker Layouts for the full list of supported loudspeaker layouts.
Object-based. The API allows for controlling the position of the object in the 3D space. Currently only single channel, point-source objects are supported.

Encoding to Ambisonics

Handling of channel and object-based audio elements is done by encoding the required input channels into an intermediate Ambisonic bed and subsequently rendering the bed using a set of FIR filters matching its Ambisonic order.
The calculation of spherical harmonic (SH) coefficients used during the encoding process is based on the Resonance Audio' s AssociatedLegendrePolynomialsGenerator class.

Ambisonics to Binaural

The binaural renderer is based on the principle of using Head Related Transfer Functions (HRTFs) and Binaural Room Impulse Responses (BRIRs) decomposed using SH coefficients.
The FIR filters were designed as a combination of anechoic and reverberant filters.
The FIR filtering (convolution) mechanism employed in this library is based on the Resonance Audio' s AmbisonicBinauralDecoder class and its dependencies.
The library uses asymmetric binaural processing, this means it employs dedicated filters for left and right ears.
Binaural filter matrices are loaded from embedded assets.
The rendering filters are selected depending on the Ambisonic order matching the input Audio Element. Channel- and object-based Audio Elements are first encoded to 3rd Order Ambisonics and then rendered using filters of the same order,
The following table summarizes the binaural filter lengths per each SH channel for Direct, Ambient, and Reverberant types at a 48 kHz sampling rate:

Filter Type	Filter Length (samples)	Sampling Rate
Direct	256	48 kHz
Ambient	8192	48 kHz
Reverberant	8192	48 kHz

Usage

C++ API

The ObrImpl class is the main interface for the binaural renderer. It provides methods to configure audio elements, process audio buffers, and retrieve information about the renderer. See obr_cli_lib.cc for an example of how to use the library. Listed below are the main methods required to configure the renderer and process audio data.

Constructor

In order to instantiate the renderer, the ObrImpl constructor needs to be called. It takes the following arguments:

buffer_size_per_channel: The size of the audio processing buffer for each audio channel.
sampling_rate: The sampling rate of the audio data.

ObrImpl(int buffer_size_per_channel, int sampling_rate);

Configuration

In order to render immersive audio using obr, the renderer needs to be configured using the AddAudioElement method. The method accepts AudioElementType enums defined within the audio_element_type.h file and BinauralFilterProfile enums defined within the audio_element_config.h file. If the second argument is not provided, the filter type defaults to kAmbient.

absl::Status AddAudioElement(AudioElementType type,
                             BinauralFilterProfile filter_type);

Audio processing callback

Once the renderer is configured, the rendering of audio data can be done using the Process method. The method uses AudioBuffer class for storing input and output audio data.

  void Process(const AudioBuffer& input_buffer, AudioBuffer* output_buffer);

Command Line Interface (CLI)

The command line interface provides a straightforward way to render audio files of certain types. The CLI is built as a standalone executable and can be run as follows:

 obr_cli
   --input_type=<1OA|2OA|3OA|4OA|7.1.4|OBA>
   --oba_metadata_file=<full path to textproto file> (required for OBA)
   --input_file=<full path to wav file>
   --output_file=<full path to wav file>
   --buffer_size=<number of samples> (optional, default is 256)
   --filter_type=<Direct|Ambient|Reverberant> (optional, default is Ambient)

Testing

The library is tested using Google Test framework. The tests are located in test subdirectories.