Media Gateway Technologies
Octasic’s media gateway products include proprietary technology for audio and video processing.
Voice Quality Enhancement & Echo Cancellation
Line echo, electrical echo or hybrid echo are different names given to the echo generated by an impedance mismatch in the analog local loop. This occurs when there is a hybrid and multiple gauges of wire, unused taps and loading coils in use. The most common source of electrical echo is the impedance mismatch caused by the 2 wire to 4 wire conversion done by the hybrid transformer, creating an electrical reflection point that is the source of echo.
Line Echo Cancellation
A Line Echo Canceller (LEC) is made of two functional parts: the Adaptive Filter (AF) and the Non-Linear Processor (NLP). The AF is a numerical filter that models the echo path, predicts the echo of the signal present on its receive path, and subtracts the estimated echo on its send path. Since the AF can only model the linear part of the echo, it will attenuate all the linear components and let through any non-linear components of the echo signal. The remaining signal, residual echo, is handled by the NLP. The NLP basically removes any signal that is below a certain threshold and replaces them with simulated background noise.
Octasic’s patented Echo Cancellation algorithm:
- Certified carrier-grade algorithm, considered a benchmark for echo cancellation
- Field hardened: deployed in wireless, wireline and VoP networks worldwide
- Seamless, transparent echo canceller with Octasic music protection
- G.168 Compliant
Acoustic echo is caused by acoustic coupling problems between a telephone's speaker and its microphone. Acoustic echo can occur in mobile phones, wireline telephones or in a hands-free set of a speaker phone. It can be caused by handset crosstalk in poor quality handsets or by echo in the environment surrounding the caller. Handset crosstalk occurs when sound coming out of the speaker is picked-up directly by the microphone. The most common source of acoustic echo is speaker phones in hands-free mode or any type of endpoint using speakers and a microphone; like a SIP endpoint running on a PC for example.
Acoustic echo is prevalent in mobile networks and is typically installed in the Mobile Switching Center (MSC). Octasic’s solutions provide both Line Echo Cancellation (LEC) and Acoustic Echo Cancellation (AEC) simultaneously in the same device, enabling high density gateways that can be provisioned for the appropriate type of call.
Acoustic Echo Cancellation
Acoustic Echo due to its mechanical and environmental sources is mostly non-linear. It is also much more time-varying than line echo since movements of the speaker or device will cause changes in the echo path. Therefore, an AEC algorithm relies much more heavily on the NLP. The echo tail for acoustic echo can also be much longer, sometimes up to half a second.
Adaptive Noise Reduction
The ubiquity of mobile phones has meant that a large percentage of calls are now made outdoors, or in noisy settings. Airports, cars, restaurants, and busy streets, all make it difficult to hear and be heard when placing a call. Adaptive Noise Reduction (ANR) serves to remove the unnecessary noise from a conversation. Removing this noise increases intelligibility, reducing listener fatigue and improving customer satisfaction.
ANR can be run on any voice call, removing the environmental noise, without clipping, distorting or otherwise modifying the speech. The algorithm runs constantly, adapting to changes in the environment.
Octasic has brought together the key ingredients necessary to develop video algorithms for tomorrow’s video networks. Whether delivering video over wired or wireless networks to HD panels or mobile devices, Octasic’s algorithms and expertise guarantee the best possible quality. A combination of high-performance devices and fine-tuned software libraries help OEMs accelerate product development.
As DSPs go multi-core to get around the power limitations of current architectures, Octasic is ahead of the pack. With its second generation of multi-core DSPs for video, the OCT2200 Series delivers unprecedented performance in a fully-programmable DSP. Processing HD resolution video is accomplished by dividing the task across multiple cores.
Current solutions use a complicated mix of DSP processors, hardware accelerators, and general purpose CPU cores in order to achieve the performance required for high-end video systems. Octasic solutions, which are powered by the most power efficient DSP core- Opus, are much simpler to program and use only one kind of processor. This reduces the number of software tools required to create and debug code, while simplifying the system design by eliminating the need for multiple steps in a video pipeline such as managing external hardware acceleration blocks.
Octasic solutions that process video completely in software allow for very flexible implementations. One of the most important things for video conferencing equipment is compatibility and inter-operability. It is important to be able to make quick fixes to the codec in software. While H.264 is the new dominant standard, backwards compatibility with video codecs include H.261, H.263, H.263+, MPEG-4. The solution must also be future-proof to support technologies such as Scalable Video Coding (H.264 SVC) and H.265.
Production Grade Codecs
Octasic offers production grade libraries for all major video codecs such as H.263, MPEG-4 and H.264. These codecs haven been deployed in many different network scenarios, leading to heavy inter-operability testing with a broad range of endpoints. These codecs have been optimized for a multi-core DSP environment, and therefore deliver excellent performance on Octasic’s DSPs.
Packet Loss Concealment
Packet loss is common on IP networks due to its fundamental “best-effort” design philosophy. When the network cannot relay a packet, it is simply discarded. The application is responsible for dealing with packet loss, either through re-transmission or concealment. When re-transmission is impractical, packet loss concealment (PLC) is the only viable approach. This most often occurs in applications that require very low delay, such as real-time communications video conferencing. Octasic’s codec libraries include algorithms that reduce the effects of packet loss. Instead of the readily-noticeable artifacts, PLC masks the loss by estimating the expected data based on previous frames.
Video encoders rely on an algorithm called Bit-Rate Controller to control the output bit-rate of the video stream. Video streams are typically Variable Bit-Rate (VBR), meaning that the amount of bandwidth required to encode a given stream can vary depending on the complexity or movement found in the source video.
Octasic has developed patent-pending algorithms for bit-rate controllers in the video over IP space. These algorithms cover the broad spectrum of real-time to streaming video applications.
The bit-rate controller must offer the highest quality video for the bandwidth that has been allocated. The following graph demonstrates Octasic’s bit-rate controller’s performance for real-time communications where delay must be minimized.
For a given video feed, the graphs below demonstrate how Octasic is able to maintain a near-constant bit rate for a real-time video conference. By keeping the bit-rate constant, we’re able to maintain a very low delay.
In order to evaluate the effectiveness of the bit-rate controller the perceived quality must also be compared. The following graphs compare the Peak Signal-to-Noise Ratio (PSNR) of each video stream. Not only does Octasic’s solution deliver a higher PSNR on average, it also maintains a more stable level of quality. This eliminates flicker, or periods of varying crispness in the image.