Independent Non-Base Layer Decoding

Description

Independent Non-Base Layer Decoding (INBLD) is an advanced feature of the 3GPP Enhanced Voice Services (EVS) codec, specified in TS 26.948. EVS is a scalable and embedded codec, meaning its bitstream is structured in layers: a mandatory base layer (BL) providing basic audio quality and one or more enhancement layers (EL) that incrementally improve quality (e.g., higher bandwidth, better stereo). Traditionally, in scalable audio coding, decoding an enhancement layer requires the successful decoding of all lower layers (including the base layer). If the base layer is corrupted or lost due to packet loss, the enhancement layers become unusable, leading to a significant drop in audio quality.

INBLD modifies this dependency structure. It allows the decoder to attempt to decode and utilize a received non-base (enhancement) layer even if the base layer for that frame is missing or corrupted. This is achieved by designing the enhancement layer coding to be partially self-contained. While the enhancement layer data is still optimally predicted from the base layer for maximum compression efficiency, INBLD includes additional information or uses coding techniques that enable a fallback decoding path. When the base layer is unavailable, the decoder uses this supplemental data within the enhancement layer packet itself, along with state information from previously correctly decoded frames, to synthesize a plausible audio signal. The output quality from decoding only an enhancement layer is lower than decoding the full scalable stream but is significantly better than having no decodable data at all (which would result in frame concealment or error).

Architecturally, INBLD involves modifications at both the encoder and decoder. The encoder must generate the enhancement layer bitstream with the necessary redundancy or independent coding elements to support the fallback mode. The decoder must implement the logic to detect base layer loss and switch to the independent decoding mode for the enhancement layer. This feature operates on a frame-by-frame basis and is a key component of EVS's robust error concealment and packet loss resilience mechanisms, which are critical for maintaining high voice quality over unreliable channels like VoLTE and VoNR.

Purpose & Motivation

INBLD was developed to address a fundamental weakness of traditional layered (scalable) codecs in lossy packet network environments like mobile IP networks (3G, 4G, 5G). In such networks, packet loss is common due to radio interference, handovers, or congestion. The EVS codec's scalable structure is efficient for bandwidth adaptation, but its strong layer dependency made it vulnerable: losing the base layer packet would nullify all enhancement layers for that frame, causing a severe and abrupt quality degradation. This was at odds with the goal of EVS to provide superior quality and robustness, especially for voice over LTE (VoLTE) and future services.

The technology solves this problem by reducing the decoder's critical dependency on the base layer, thereby "smearing" the impact of packet loss. Instead of an all-or-nothing scenario, INBLD provides a graceful degradation path. When the base layer is lost, the network or receiver may still have the enhancement layer packet (which might have taken a different path or be in a separate RTP packet). INBLD allows the system to salvage this data to produce usable audio, preventing a complete breakdown in perceived quality. This is particularly important for high-bitrate stereo music or high-quality voice calls where the enhancement layers carry significant perceptual information.

The motivation for INBLD came from the evolving use cases for the EVS codec, which was designed not just for narrowband voice but also for high-quality music streaming, conference calls, and immersive communications. These applications demand higher reliability and quality consistency. By introducing INBLD in Release 13, 3GPP enhanced the EVS codec's error resilience without sacrificing its core scalable efficiency, ensuring it could deliver on the promise of high-quality audio services across the variable conditions of mobile broadband networks.