Spectral band replication
It is used to enable encoding methods for lossy compression of audio files at low bit rates (typically 24 kbps to 64 kbps) of acceptable quality. For this purpose, frequency components above a cut-off frequency (for example between 4 and 8 kHz) are no longer coded directly, but are regenerated using control signals from directly coded middle and lower signal ranges using similar-sounding signals.
This is made possible by two properties:
- The human hearing becomes less and less demanding at higher frequencies. On the one hand, phase positions are no longer recognized from around 2 kHz, on the other hand, the difference in sound between sine tones and narrowband noise at high frequencies becomes smaller and smaller.
- The spectrum of many sounds has characteristic envelopes; There is a clear correlation between the envelope in the low-frequency and high-frequency range. Overtones are continued periodically, the temporal envelopes of medium-frequency and high-frequency tones usually look very similar.
To synthesize the supplementary high-frequency signal, a frequency-transformed representation of the signal is usually used. In this way, the frequency bands can be replicated in a simple manner. There are various methods for doing this. In the sub-band method, suitable sections of the decoded low-frequency partial signal are selected for different sub-bands of the supplementary signal to be synthesized and combined at an adapted volume. The encoder determines the parameters for controlling the reconstruction process.
Similar to many other coding methods ( e.g. intensity stereophony ), this coding method can hardly be used for transparent codings or is a hindrance, since the method by its nature synthesizes signal components independently of the corresponding component of the original signal. The main application is in the coding of medium data rates with medium quality.
As an alternative method that is not encumbered by patents, CELT , for example, uses a technique known as frequency band folding, which is supposed to achieve similar results and has significantly lower implications for codec latency and complexity (computational effort).
- Audio format mp3PRO , 64 kbps, conventional coding of the spectral range between 0 and 8 kHz, SBR between 8 and 16 kHz.
- Audio format MPEG-4 High Efficiency Advanced Audio Coding (HE-AAC, AAC +, AACplus v1, AACplus v2), 32 kbps, conventional coding of the spectral range between 0 and 5 kHz, SBR between 5 and 15 kHz.
- Audio format Enhanced Audio Codec 2.0
- patent WO9857436 : Source Coding Enhancement using Spectral Band Replication. Published on December 17, 1998 , inventors: Liljeryd, Lars Gustaf.
- Archived copy ( memento of the original from September 24, 2015 in the Internet Archive ) Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice.