Variable Length Code

Description

Variable Length Coding (VLC) is a fundamental lossless data compression algorithm based on information theory principles, specifically Huffman coding. The core idea is to assign shorter binary codewords to source symbols (e.g., parameter indices, transform coefficients) that have a higher probability of occurrence, and longer codewords to less probable symbols. This reduces the average number of bits per symbol, achieving compression without losing information. The codebook, which maps symbols to variable-length bit patterns, must be known by both the encoder and decoder. The decoder parses the incoming bitstream uniquely, as no codeword is a prefix of another (the prefix-free property), allowing for unambiguous decoding.

Within 3GPP specifications, VLC is primarily applied in the source coding stage of speech and audio codecs defined in the 3GPP TS 26-series, such as the Adaptive Multi-Rate (AMR) codec and the Enhanced Voice Services (EVS) codec. These codecs operate by analyzing the input speech signal to extract parameters like Line Spectral Frequencies (LSF), pitch lag, and algebraic codebook indices. The statistical distribution of these parameters is not uniform; certain values occur far more frequently than others. VLC is employed to efficiently encode these parameters for transmission. For instance, after quantization, an index representing an LSF vector is fed into a VLC module, which outputs a compact, variable-length bit sequence. This bitstream is then multiplexed with other encoded parameters into a single frame for transmission over the radio interface.

The implementation involves designing optimized codebooks tailored to the probability distribution of the parameters for different operating modes (e.g., different bitrates, background noise conditions). In advanced codecs like EVS, multiple VLC tables may be switched dynamically based on signal characteristics to maximize coding efficiency. The decoding process in the receiver involves a bit-by-bit parsing of the received frame, using the same VLC table to reconstruct the parameter indices, which are then used to synthesize the speech signal. This efficient entropy coding is a key reason why modern codecs can provide high-quality speech at very low bitrates, directly impacting spectral efficiency and user capacity in mobile networks.

Purpose & Motivation

VLC was developed to address the need for efficient digital representation of information, minimizing the number of bits required for storage or transmission. Prior to such entropy coding techniques, fixed-length coding was used, which wasted bits on representing low-probability events with the same length as high-probability ones. Claude Shannon's information theory established that the optimal code length for a symbol should be proportional to the negative logarithm of its probability. VLC algorithms like Huffman coding provide a practical method to achieve or approximate this theoretical limit, solving the problem of redundant bit allocation in digital communication systems.

The incorporation of VLC into 3GPP speech codecs was motivated by the stringent bandwidth constraints of early cellular systems (2G/3G) and the continuous drive for higher capacity. Voice has always been the primary service, and efficient coding directly translates to more simultaneous calls per cell. By applying VLC to speech codec parameters, engineers could significantly reduce the average bitrate of the encoded speech frame without degrading perceptual quality. This allowed for the development of multi-rate codecs like AMR, which could operate efficiently under varying channel conditions. The evolution to wideband and super-wideband audio (EVS) further leveraged advanced VLC techniques to maintain high quality at competitive bitrates, ensuring efficient use of both circuit-switched and packet-switched (VoLTE, VoNR) network resources. Thus, VLC is a cornerstone technology enabling the cost-effective, high-quality voice services that underpin mobile telecommunications.

Key Features

• Lossless compression based on symbol probability (entropy coding)
• Assigns shorter codewords to more frequent symbols for bitrate reduction
• Uses prefix-free codes (e.g., Huffman codes) for unambiguous decoding
• Integral part of 3GPP speech/audio codecs (AMR, EVS) for parameter encoding
• Dynamically selectable codebooks adapt to different speech statistics
• Directly improves spectral efficiency of voice services in mobile networks

Evolution Across Releases

Defining Specifications