Advanced Audio Coding Long Term Predictor Object Type

Description

Advanced Audio Coding Long Term Predictor (AAC-LTP) is a specific audio object type defined within the MPEG-4 AAC standard and adopted by 3GPP for use in mobile telecommunications. As an object type, it represents a particular configuration or profile of the AAC codec, specifying which coding tools are used. The core of AAC-LTP is the integration of a Long Term Predictor into the standard AAC transform coding framework. AAC itself is a perceptual audio codec based on a modified discrete cosine transform (MDCT) that converts time-domain audio signals into a frequency-domain representation, allowing for efficient quantization based on psychoacoustic models to discard inaudible information.

The Long Term Predictor (LTP) is the key distinguishing feature of this object type. Unlike the standard AAC which may use other prediction tools like the Temporal Noise Shaping (TNS) or rely solely on intra-frame coding, LTP introduces a form of long-term prediction across multiple audio frames. It works by analyzing the periodic or quasi-periodic components in the audio signal, which are common in voiced speech and certain musical tones. The predictor identifies redundancy between distant samples (over a longer time span than short-term predictors) and encodes this redundancy more efficiently, reducing the bitrate required for a given level of perceived audio quality or improving quality at a fixed bitrate.

Architecturally, within the 3GPP system, AAC-LTP is implemented as part of the media codec layer in the User Equipment (UE) and network elements like the Media Gateway (MGW) or Multimedia Resource Function Processor (MRFP). It operates on audio frames, typically of 20ms duration. The encoding process involves: 1) Time-frequency transformation via MDCT, 2) Psychoacoustic modeling to determine masking thresholds, 3) Application of the Long Term Predictor to exploit inter-frame redundancy, 4) Quantization and entropy coding of the spectral data and predictor parameters. The decoder reverses this process, using the received LTP parameters to accurately reconstruct the predicted signal components, thereby synthesizing the output audio.

Its role in the 3GPP network is primarily within the bearer and service layer for audio delivery. It is a supported codec for circuit-switched and packet-switched voice services (e.g., in early EPS deployments for VoLTE fallback scenarios), multimedia messaging (MMS), and streaming services. The inclusion of LTP makes it particularly robust and efficient for speech-dominated content compared to the AAC Main or AAC Low Complexity (LC) object types, offering a good balance between computational complexity and compression performance. It is defined in 3GPP TS 26.401, which specifies the general audio codec audio processing functions.

Purpose & Motivation

AAC-LTP was created to address the need for a high-quality, efficient, and robust audio coding standard suitable for the constraints of mobile networks. Prior to its standardization in 3GPP Release 8, mobile audio services relied on codecs like AMR-NB/WB for speech and simpler AAC profiles for general audio. These codecs were often optimized for a specific type of content (speech vs. music) and could be inefficient or provide lower quality for mixed content. The motivation was to leverage the advanced compression of MPEG-4 AAC while enhancing its performance for the periodic signals characteristic of speech, without requiring the very high complexity of the AAC Main profile's backward-adaptive prediction.

The Long Term Predictor specifically solves the problem of efficiently encoding the pitch periodicity in voiced speech segments. Standard AAC, while excellent for music, does not explicitly model this long-term correlation, potentially leading to higher bitrates for speech or artifacts. By integrating LTP, AAC-LTP achieves better speech quality at lower bitrates compared to AAC-LC, making it more suitable for voice-centric applications and services where bandwidth is limited. Furthermore, its design considers the error resilience needs of wireless channels; the predictor structure can help mitigate the impact of frame losses to some degree, as the decoder can use past correctly received frames to aid prediction.

Historically, its introduction in Release 8 coincided with the expansion of 3GPP services beyond basic voice into richer multimedia. It provided network operators and device manufacturers with a standardized, royalty-bearing but technically advanced audio codec option that could handle both speech and audio with good efficiency, supporting the evolution towards converged IP-based multimedia services (IMS). It addressed limitations of previous codecs by offering a wider audio bandwidth and better quality for music than pure speech codecs, and better speech efficiency than basic AAC profiles, thus serving as a versatile tool in the 3GPP audio codec toolbox.

Key Features

• Integrates a Long Term Predictor (LTP) tool into the AAC framework for enhanced compression of periodic signals
• Defined as a specific MPEG-4 Audio Object Type (profile) within the 3GPP standards
• Provides improved speech coding efficiency and quality compared to standard AAC Low Complexity profile
• Supports a wide range of bitrates and audio bandwidths suitable for mobile applications
• Designed with considerations for robustness in error-prone wireless transmission environments
• Enables high-quality audio for services like voice calls, messaging, and streaming in 3GPP networks

Evolution Across Releases

Defining Specifications