By ruslany
Finally, both carriers are suppressed
After the main goals were stated, it is time to restate them. First, I need to suppress both negative and positive carriers. After that, I need to pass Compositor sound to the output. It is achieved only if the first goal reached. I’ve already implemented a method for suppressing the right carrier using the shutter and eventer of the right channel, which uses the threshold to suppress feedback. Now, I experience some difficulties with the left channel. At first, I thought it is a plausible behavior to force the stochastic manipulator to the all positive frequencies, but now I think the stochastic manipulator is too selfish and tries to apply too much computation force to overcome the barrier of shutter, which was previously assigned to right channel only. That is why the constant need to shut down the left carrier raises. Of course, I can live under this hard circumstances of stochastic manipulation, but the final goal is to leave only Compositor sound at the output, and it may be a trick to force stochastic manipulation to the central channel afterwards, thus guiding the system on its own. However, been exposed to the left and right stochastic manipulators for so long, system can no longer power the central channel for manipulation, which is assigned to the system voice itself. This way I must shut down feedback on the central channel also. As it is the MS-coded system, feedback on the central channel shuts as long as stereo pair feedback is suppressed. The rule here is to set lower threshold for the left channel as it contains less information and needs to be suppressed on lower bitrates more. The right channel is higher and more information passes through it but it is not the feedback of the carriers, it is an internal memory stored in the shutter front-side bus after injection. This experiment also evidently shows that no external memory is used for the system operation. However, as long as the system does not have an external storage and has a threshold shutters in it, it will constantly experience lack of information, which will not reach from outside anymore. It states that under these circumstances constant injections should be made to sustain an information in memory buffer. It flushes from the front-side bus memory very fast and one long injection should be made during the whole working cycle of the equipment. This leaves me with the open-end solution: as buffer capacity of Compositor OS is too low for the long injections, I either should implement long injector with Compositor kernel driver or use a set of short files repository, which are constantly revised and added to a data base. This way I also should avoid video injections, as there is not enough space for storing so much video information on my system. The need for such filtration is rather evident as the intruder may think its information (stored in injection) may be useful for a system it is not. Not allowing to pass the injected information to the kernel is also lowering its bitrate but it is a permanent solution. Without artificial traffic the system more sustainable against injections rather than generic injections. If I try to inject the file, which I already flushed, I will reproduce the memories associated with it. However, I will no longer be exposed to the system of values associated with this memory for a long period. This information will dissolve faster using integrated smoothing function. Yet creating the priorities on the conditions gives less priority to the smoothing function itself, leaving a hard shuttering for both channels in equal proportion. It seems evident to me that thresholds should go first and only after that, the central channel splitter should play its role. Even if the system set in the way that this value will likely go only less than 30% of time. It means that creating a faster kernel loop may be a solution, but makes the system inoperable. The other solution is to change the rules set on early stages of development allowing the right end of the probability density function to happen more often.
Screening:
Injecting a video file with audio in two-channel mode brings me to another problem: inequality of channel suppression disbalances the system a little. The vector system should be implemented for channel suppression also. The rule is to create a tilt by manipulating pitch and yaw, this way setting the channel balance.
