Object-Based Audio

对象式音频 is a method of production and transmission that represents the sound in a program as one or more audio objects, to which attributes such as position, size, motion, gain, priority, or interaction are assigned. The renderer on the playback end assigns objects to speakers or converts them into binaural signals for headphones based on this audio data, metadata, and the actual playback environment.

The main difference from fixed-channel audio lies in the program description layer. In 5.1- or 7.1-channel mixes, each channel is predefined to correspond to a specific speaker position; in object-based programs, dialogue, effects, or other sound sources can be described as independent objects, allowing the same master to adapt to different numbers and layouts of speakers. Actual releases often employ a hybrid structure: content such as music and ambient sounds first forms a fixed channel bed, while a small number of sounds requiring precise positioning or movement are treated as objects.

An object itself typically consists of a segment of mono or multichannel audio along with time-varying metadata. The renderer must handle coordinate systems, speaker capabilities, room layouts, object diffusion, loudness, and downmixing rules. When the number of speakers is limited, multiple objects may be mapped to adjacent speakers; for headphone playback, directional cues are created using methods such as head-related transfer functions. Object coordinates are not traditional audio channels that can be directly routed to a single speaker.

对象式音频 is not a standalone compression codec. Audio samples can be encoded using lossless or lossy methods, and metadata can be encapsulated according to different broadcasting, cinema, file, and disc systems. Systems such as Dolby Atmos, DTS:X, and MPEG-H 3D Audio all utilize the concept of objects, but their respective bitstreams, metadata, rendering constraints, and compatibility layers differ; therefore, “对象式音频” cannot be treated as a synonym for any specific commercial format.

The ITU’s audio definition model classifies direct speaker, matrix, object, higher-order Ambisonics, and binaural audio as distinct types. Object-based and scene-based audio are also distinct: scene-based audio represents the spatial basis functions of the entire sound field and is decoded as a whole during playback; object-based audio, on the other hand, preserves several sound sources that can be individually identified and controlled. A program may use a combination of these representations.

The final listening experience depends on how many objects are preserved in the distribution stream, the decoding level supported by the playback device, the renderer, and the actual speaker layout. The presence of Atmos or DTS:X logos on a player merely indicates that the corresponding format has been recognized; it does not determine the number of objects, whether objects are used throughout the entire content, or whether all sound sources have discrete height positions.