Siren (codec)

Siren is a lossy audio compression data format based on a frequency transform.

Siren comes in three variants that have different widths frequency reproduction: Siren7, Siren14 and Siren22 with appropriate reproduction of frequencies up to 7, 14 or 22 kHz. This bit rates of 16, 24, 32, 48 or 64 kb / s are possible for monophonic reproduction, for stereo twice for the required. It establishes an algorithmic delay ( latency ) of 40 milliseconds.

For the 7 -kHz variant nachkonstruierter for the Instant Messenger aMSN free codec exists in the form of a program library called " libsiren ".

Technology

The method divides the signal into overlapping blocks of 20 milliseconds duration. For this, the quantized mean-square value of the amplitude is explicitly transmitted, and the residual signal described with frequency coefficients from a modified discrete cosine transform, a variant of the discrete cosine transformation of type 4, which are grouped into 500 Hz wide sub-bands, and are quantized. The values ​​are packed with a kind of Huffman coding.

History

Development of Siren codecs began at the company PictureTel Corp.. for telephony purposes. This developed the PT724 and PT716 audio formats that could at data rates of 24 or 16 kb / s transmit audio signals with frequencies up to 7 kHz ( broadband). This was followed by the development of " PT716plus ", which is the direct precursor of the Siren codecs. With the takeover of the manufacturer's rights to the technology fell in October 2001 at Polycom. Part of the technology (mainly Siren7 and Siren14 ) was established in 1999 standardized by the ITU- T G.722.1 and as released in the course of this conditional free use. The G.719 adopted in 2008 builds on Siren22.