The following is the list of Speech Decoders and Encoders implemented on ARC processors. As always, our team has the capabilities to quickly port these solutions on other popular DSP and RISC processors.
|
|
G.711 Codec |
|
G.711 is an ITU-T standard
for speech codecs that provides toll quality speech at
64kbps using either A-law or mu-law PCM methods. While
mu-law is generally used in North America and Japan,
A-law is used in Europe and the rest of the world. This
falls in the category of waveform speech coders which
try to compress the waveform as it sample by sample.
The Incube G.711 codec is very compact and efficient
implementation. Data
Sheet |
|
G.726 Codec |
|
G.726 is an ITU-T ADPCM
speech codec standard covering the transmission of voice
at rates of 16, 24, 32 and 40 kbps. These four bit-rates
are often referred to by the bit size of a sample, which
are 2 bits, 3 bits, 4 bits and 5 bits respectively, since
the sampling frequency is fixed to 8 kHz. This codec
falls in the category of waveform speech codecs. Data
Sheet |
|
G.729 AB Codec |
|
G.729 is an ITU-T speech
codec that adopts Conjugate Structure Algebraic Code
Excited Linear Prediction (CS-ACELP) scheme. It is a
parametric speech codec that operates at 8 kbps and finds
applications in telephony over packet networks and Voice
over IP. G.729A is the reduced complexity version of
G.729, which marginally sacrifices the speech quality
to achieve reduction in complexity. The annex B of G.729
is a silence compression scheme which uses a VAD module
to detect voice activity in the signal, a DTX module
which decides on updating the background noice parameters
for non-speech frames, and a CNG module that generates
the required comfort noise akin to the old analog hiss.
Incube G.729 codec is an implementation of the low-complexity
version of G.729 with Annex A and B. Data
Sheet |
|
AMR-NB Codec |
|
AMR-NB or simply AMR is
a patented audio data compression scheme optimized for
speech coding. AMR uses the Algebraic Code Excited Linear
Prediction (ACELP) in addition to DTX, VAD and CNG to
reduce bandwidth usage during silence periods. It uses “link
adaptation” to select one of eight different bit-rates
based on link conditions. It was adopted as the standard
speech codec by 3GPP and is widely used in GSM and UMTS. Data
Sheet |
|
AMR-WB Codec |
|
Adaptive Multi Rate – WideBand
is a parametric speech codec that provides excellent
speech quality due to the use of wider speech bandwidth
of 50-7000 Hz as compared to narrow band codecs that
which are optimized for POTS wireline quality of 300-3400
Hz. AMR-WB uses the Algebraic Code Excited Linear Prediction
(ACELP). The sampling frequency used is 16 kHz as opposed
to 8 kHz used for narrow-band codecs. AMR-WB is codified
as G.722.2, an ITU-T standard speech codec.
Incube AMR-WB codec has been heavily optimized to
enable its use even in low-power applications. Data
Sheet |
|
AMR-WB+ Codec |
|
AMR-WB+ is an extension
of AMR-WB. The main improvement is the use of transform
coding additionally to ACELP. This greatly improves the
performance of this codec for audio signals. Automatic
switching between transform coding and ACELP provides
both good speech and audio quality with moderate bit-rates.
It adds support to stereo signals, higher sampling rates
and a wide range of bit-rates. While doing all this,
AMR-WB+ manages to be backward compatible to AMR-WB.
Building on its highly efficient AMR-WB codec implementation
and expertise in audio codec development, Incube AMR-WB+
codec implementation provides a very attractive MHz
proposition for the end customer using AMR-WB+ codec. Data
Sheet |
|
EVRC-B Codec |
|
Enhanced Variable Rate Codec-B (EVRC-B)
is a 3GPP2 standard 4GV codec for cellular phones in
CDMA2000 systems. EVRC-B is an extended version of EVRC
which is a Relaxed Code Excited Linear Prediction
(RCELP) based speech coding algorithm. In addition to
RCELP, EVRC-B uses Prototype Pitch Period (PPP) approach
for coding of stationary voice frames and Noise
Excitation Linear Prediction (NELP) for efficient coding
of unvoiced or noise frames. Thus it provides lower data
rates compared to EVRC, as it includes 1/4 rate frames,
null frames and erasure frames compressing each frame of
data into one of the four bit rate frames of 9.6, 4.8,
2.4 and 1.2kpbs. EVRC-B has operation points (0,1 and 2)
facilitating superior flexibility in rate assignment
supporting average bit rate in the range of 9.3 - 4.8
kbps. |
|
Incube's EVRC-B codec, with all the above features, is
efficiently optimized for low MHz and low power
real-time operation. |
|
Speex Codec |
|
Speex is
an open source lossy speech codec for mainly VOIP
applications. Incube's Speex codec is a highly optimized
codec supporting narrowband, wideband and ultra-
wideband (for 8, 16 and 32kHz sampling rates
respectively). It supports both variable bit rate mode
(VBR) and average bit rate mode (ABR) operations
resulting in a very low bit rate ranging between 2 – 44
kbs. Its features include voice activity detection
(VAD), comfort noise generation (CMG) and discontinuous
transmission (DTX).
Data
Sheet |
|
SILK Codec |
|
SILK is CELP based speech algorithm used for packet based voice communications. Silk provides scalability in several dimensions to target a range of operating environments. The features of SILK codec make it useful in different network conditions are control of bit rate, packet rate, resilience to packet loss and use of discontinuous transmission (DTX). SILK codec can be set at different complexity levels to suite different processor capabilities. SILK decoder integrated with sample rate converter, it converts output signal to 8, 12, 16 or 24 KHz frequency based on the user option. Though SILK’s main applications are for internet based voice communications, its super-wideband capabilities (24 kH) make it a potential candidate for internet based audio transmission.
Incube's implementation of SILK codec supports all the above features and is highly optimized for low MHz and smaller memory footprint.
|
|
 |
Speech
Processing: |
|
|
Acoustic Echo
Canceller (AEC): |
|
Acoustic Echo
Canceller (AEC) is a technique used to suppress
acoustic echo in teleconferencing and hands-free
telephony applications. Incube's AEC consists of
efficient adaptive filter to provide over 40dB echo
suppression up to echo tail-lengths of 64 ms. It
operates in full-duplex mode with a robust double talk
detector with high rate of double talk detection and
negligible false alarms. It operates in both narrowband
and wideband modes (for sampling frequencies of 8 and 16
kHz respectively). The Incube's AEC software provides
a complete solution with features such as noise
suppression, voice activity detection (VAD), comfort
noise generation (CMG) and automatic gain control (AGC).
Data
Sheet |
|
Automatic Gain
Control (AGC) |
|
Incube's
Automatic Gain Controller (AGC) is well suited
for telephonic applications with smooth gain variations
and well calibrated thresholds for noise detection (to
avoid unnecessary amplification of noise during
no-speech periods) and maximum gain (to avoid saturation
of speech).
Data
Sheet |
|