About AAC
Advanced
Audio Codec 
C55x
Encode, C55x
Decode, LC
- C64x Enc/Decode
Advanced audio coding (AAC) is a MPEG standard covering a family of audio encoding schemes. AAC improves upon MP3 in several ways to achieve considerably improved perceived sound quality at lower bitrates. AAC is also more flexible, providing codec developers more freedom in bitrates, sampling rates, frame lengths, and encoding approaches than is possible in the context of MP3. AAC also uses the new MP4 container to store metadata regarding the encoded audio.
Improving upon MP3, AAC provides for higher resolution audio with sampling from 8 kHz to 96 kHz sampling, 48 possible audio channels, more compact encoding and less-computationally intensive decoding. Unlike MP3's hybrid filter bank, AAC uses the modified discrete cosine transform (MDCT) with up to 1024-point windows, which can be switched to shorter 192 to 128-point windows during transients. The longer coding windows improve the coding efficiency of AAC. AAC employs two central coding techniques to compress audio data - portions of the signal that are not perceptually relevant or audible are removed, and redundant data in the coded signal itself are also removed. AAC employs Temporal Noise Shaping (TNS) to control where, in time, quantization noise will occur by encoding filtering coefficients with the signal. Perceptual Noise Substitution (PNS) allows AAC to encode signal noise with a more compact representation instead of coding the more complex waveform.
While AAC is a significant improvement over MP3, there are some significant limitations as compared to MP3. First, AAC is more complex than MP3 in standard form, requiring more processing power to decode than a similar MP3 decoder. Second, AAC is available in a wide range of formats and profiles, few of which are used consistently. While most decoders support the Low Complexity profile, many do not support some of the more obscure profiles such as AAC-HE (High Efficiency). This can result in fully-compliant audio files that simply will not work with fully-compliant decoders that do not recognize the encoding profile. Third, AAC is tied to significantly more patents than MP3, potentially encumbering its future portability.
AAC is available in a number of sub-formats, referred to as profiles. One such profile, referred to as AAC-LC for Low Complexity, is the simplest profile used in most codecs and implementations of AAC. AAC-LC also happens to be the most widely supported AAC profile for exactly these reasons. Using a mere 64 kbps per audio channel, AC-LC can deliver compressed audio with such high quality that it is virtually indistinguishable from the original uncompressed audio.
Another profile, referred to as AAC-LD for Low Delay, is ideally suited to systems requiring high quality communications. With an MP3 encoder/decoder chain the algorithmic delay can exceed 100 ms, impeding natural communication. Alternatives to MP3 for low delay coding can improve upon the delay in MP3, but only by focusing on speech and ignoring the frequency range and acoustic characteristics of music. AAC-LD combines the perceptual audio coding qualities of MP3 with low delay and multi-channel support. The algorithmic delay of AAC-LD is a mere 20 ms, allowing for more natural communication, making it ideal for telephone, cell phone, and video-conferencing systems.
Contributed by Aaron R. Herrmann
College of Engineering
Villanova University
