Codec2: speech compression up to 2400 bps (under LGPL2)
The first alpha version of the free voice codec Codec2 , which is designed for ultradense encoding of speech on narrow channels, has been released. For example, it compresses 3.75 seconds of distinguishable speech with just 1050 bytes. Communication quality is ensured by a stream of 2550 bps.
A Linux / gcc simulator (c2sim) has been released to test various encoding options, a separate encoder (c2enc) and a decoder (c2dec) with an interface via the command line, they work about 10 times faster than real-time on a modern PC. Download the source code and documentation here .
Such superdense compression can be useful not only for amateur radio, but also for various military missions, satellite communications, and encrypted devices. At present, the US Army uses the MELP (Mixed Excitation Linear Prediction) coding standard, but this is the intellectual property of Texas Instruments (2400 bps), Microsoft (1200 bit / s transcoder), Compandent (2400, 1200, 600 bps) , Thales (600 bps) and AT & T (noise preprocessor).
The developer of Codec2 is David Rowe (David Rowe), who previously had a hand in the creation of Speex . He set the task to achieve voice transmission in communication quality in a stream of 2400 bps and lower, that is, to make a free alternative to MELP.
')
The problem of the lack of free codecs in a range of up to 5 kbps was raised by Bruce Perens last year. He contacted the Speex developers and invited them to study the situation.
Codec2 is based on scientific papers from the 60s and 80s and does not seem to fall under existing patents. Sinusoidal speech coding was first mentioned in 1984, and Rowey himself described in detail the techniques of harmonic sinusoidal coding in his 1997 scientific work . The codec is published under the LGPL2 license.
On the samples below, you can compare the coding quality of Codec2 v0.1 (2550 bps) and MELP (2400 bps).
Male voice:
Original
Codec2 v0.1 (2550 bps)
MELP (2400 bps)
Female voice:
Original
Codec2 v0.1 (2550 bps)
MELP (2400 bps)
Of course, David Rowey plans to further finalize Codec2 significantly. You can follow the latest news in the mailing list .
A Linux / gcc simulator (c2sim) has been released to test various encoding options, a separate encoder (c2enc) and a decoder (c2dec) with an interface via the command line, they work about 10 times faster than real-time on a modern PC. Download the source code and documentation here .
Such superdense compression can be useful not only for amateur radio, but also for various military missions, satellite communications, and encrypted devices. At present, the US Army uses the MELP (Mixed Excitation Linear Prediction) coding standard, but this is the intellectual property of Texas Instruments (2400 bps), Microsoft (1200 bit / s transcoder), Compandent (2400, 1200, 600 bps) , Thales (600 bps) and AT & T (noise preprocessor).
The developer of Codec2 is David Rowe (David Rowe), who previously had a hand in the creation of Speex . He set the task to achieve voice transmission in communication quality in a stream of 2400 bps and lower, that is, to make a free alternative to MELP.
')
The problem of the lack of free codecs in a range of up to 5 kbps was raised by Bruce Perens last year. He contacted the Speex developers and invited them to study the situation.
Codec2 is based on scientific papers from the 60s and 80s and does not seem to fall under existing patents. Sinusoidal speech coding was first mentioned in 1984, and Rowey himself described in detail the techniques of harmonic sinusoidal coding in his 1997 scientific work . The codec is published under the LGPL2 license.
On the samples below, you can compare the coding quality of Codec2 v0.1 (2550 bps) and MELP (2400 bps).
Male voice:
Original
Codec2 v0.1 (2550 bps)
MELP (2400 bps)
Female voice:
Original
Codec2 v0.1 (2550 bps)
MELP (2400 bps)
Of course, David Rowey plans to further finalize Codec2 significantly. You can follow the latest news in the mailing list .
Source: https://habr.com/ru/post/104662/
All Articles