Problems of automation of the broadcasting and production complex

Kuznetsov Svyatoslav Kuznetsov Svyatoslav Sergeevich, Technical Director of San Porto and Torre Ricca TV channels Torre Ricca Limited Liability Company, Moscow 115533, Russia, Moscow, Andropova avenue, 22 svyat.smith@yandex.ru

Introduction

The broadcasting and production complex, which is a combination of technical means and software that ensure the effective operation of companies in the field of television and radio broadcasting, is the basis of the workflow and an example of the use of progressive technological methods of television production. Such complexes make it possible to cover all its stages – from filming to automated broadcasting of television programs (they have been actively used in the television industry for the past ten years). At the same time, taking into account global digitalization and increasing competition in the market, the television industry companies face the task of constantly improving broadcasting and production complexes. Such improvement is largely provided by new technologies, some of which are aimed at automating broadcasting processes.

Materials and methodsWhen writing the article, a comprehensive approach to the problems of research was implemented, in particular, a list of issues related to the introduction of artificial intelligence technologies into the broadcasting sphere was considered.

Special and general research methods were used to identify problems of automation of broadcast production complexes. Among the general methods used in the study, methods of systematic, quantitative and qualitative analysis, synthesis, as well as the formal logical method and the method of theoretical generalization should be distinguished.

The use of various research methods made it possible to achieve the validity of the obtained scientific conclusions.Results and their discussion

Automation of broadcast production complexes is not a new concept in itself and has been around for many years.

However, questions about the expediency and problems of automation of broadcast production complexes allow us to remain relevant in the field of the television industry. In particular, there are questions about improving the already existing broadcast production complexes in terms of automating the process of creating informative content, while maintaining high production performance.

For example, the traditionally accepted scheme of creating a television program with the involvement of highly qualified specialists, such as: director, technical director, graphic operator, assistant producer, sound engineers, lighting directors, can be replaced by an automation system. In other words, this means that instead of an entire team managing all the equipment of broadcast production complexes and their software, only one person will control all processes using special software that provides automation of processes.

At the same time, considering the problems of automation of broadcast production complexes from a theoretical point of view, it is necessary to focus on whether automation can always be an effective solution to the problem in the field of broadcasting.

The answer to this question needs to be compared with each specific problem that automation is trying to solve, as well as with what is required for each production process in the field of broadcasting and how it fits into the existing work processes and culture of a particular company.

In the specialized literature devoted to the development of automation processes in the field of television and radio broadcasting, as a rule, when the prospects of automation of broadcasting processes and the advantages of such automation are investigated, it is indicated that automation of broadcast production complexes is aimed at saving financial resources of TV and radio companies through one-time investment and further savings on high operating costs ^{[1, 3Such savings are achieved, among other things, by reducing the number of employees who control the process of broadcasting and releasing programs [6]}. Also, in the specialized literature, one can find an indication that automation in broadcasting allows using specialists with lower qualifications to control individual processes, in particular, the program producer can be replaced by any specialist who can switch programs using equipment ^{[2, 4, 7]}.

Automation is considered as a promising direction in the field of broadcasting, since it allows you to create more accurate, consistent and higher production standards by performing the same sequence by programmed machines, eliminating the human factor of any potential errors ^{[5, 7, 9]}.

It is impossible to agree with this opinion in general, since automation of production processes will be an effective and economical technology only if the software is properly configured and the prospects for automation are carefully calculated. Erroneous, according to the author of this article, is the indication available in the specialized literature that any employee with any education can manage the system and program. So, for example, if the concept of the program does not change within the framework of the broadcast, then there is really no need to involve a specialist with any special knowledge in the field of broadcasting to switch the transition button to the next segment. However, the above does not apply to those programs that are broadcast live, since the fullness of live broadcast programs can depend on many factors, such as the latest news, reports from the event sites, direct connection of correspondents from places where good communication quality is not provided, and so on. All of the above can cause a number of technical problems in the production process of the program, which cannot be predicted using simple automation tools. 

In the traditional workflow, the producers of the program can change its order and inform the director about it, who will report the problems to the film crew. When, as part of automation, the software replaces the film crew, the director must make sure that the appropriate encoding commands are set, aimed at the output of subsequent segments on the air. In an uncertain situation, it is easier to react quickly to any events with the help of the command that ensures the release of the program than to enter the appropriate command into the software. In addition, the person entering the command into the software must have certain technical skills and knowledge to perform all settings and encodings in accordance with the broadcast standard before going on the air.

The issue of automation costs is also very relevant. So, for example, if any additional equipment such as a wearable Steadicam stabilization system is needed during the production of a program for the production of a shooting program for which work is carried out in constant motion, then the costs of the corresponding equipment are justified. But if only a sitting presenter is being filmed as part of the program, then spending on Steadicam is unnecessary, since you can use pre-saved pictures.

The above-mentioned practical problems are not the only ones that arise within the automation of broadcast production complexes, but only based on the problems considered, it can be concluded that automation in the field of broadcasting requires additional research. Automation systems, of course, provide the ability to perform individual tasks in the field of broadcasting, but so far they are not so perfect that it is possible to talk about replacing specialists in the field of broadcasting with automated programs. All existing solutions in the field of broadcast automation are reduced to two possible options – a new automation system or a semi-automatic system using existing equipment ^[10].

The second option from the point of view of ensuring organizational goals in modern conditions seems more acceptable. Thus, in the presence of an already functioning broadcasting and production complex, the refinement of technical solutions through automation is possible without prejudice to production teams.

Each project that is being developed in the field of broadcasting differs from another project at the micro level, so the employees involved in the project can choose the amount of automation that they want to include. But at the macro level, the results of projects are often very similar. By combining two components – the macro level and the micro level, two efficient systems can be obtained that can scale according to the needs of television production and business. And such efficiency is achieved precisely due to the fact that the specifics of the work are analyzed and taken into account, and only then the software necessary for automation is selected and only to the extent that makes it easier for the employee to prepare a specific program.

In the future, taking into account the global digitalization of all production processes, the boundaries between editorial and technical production within broadcasting will continue to blur. And in this regard, it is possible to identify those areas of automation of broadcast production complexes that will really simplify the work of specialists and save organizational resources at the same time. According to the author of this article, among the promising areas of automation is voice control, which is based on artificial intelligence based on various learning algorithms for artificial neural networks.

In particular, voice control can replace a number of mechanical actions performed by the operator with simple voice control, and command recognition and actuation will be carried out according to a predetermined algorithm for a trained neural network.

Today, artificial neural networks are an integral part of most data mining systems that can influence the processes occurring in various systems due to the relationship between data about a specific event and the possibility of predicting similar events in the future ^[8].

Neural network training technologies are diverse, but for the purposes of voice control in the field of production broadcasting, the most promising technology is deep learning, due to which it becomes possible to solve a large range of robotic tasks in the fields of perception and control. The deep learning capabilities of neural networks allow you to analyze complex disparate data and build complex concepts from simpler ones. All deep learning algorithms use different types of neural networks to perform certain tasks. Deep learning algorithms train machines based on examples that are set as learning parameters.

For example, we can have three functions chained together to form. This chain of structures is the most commonly used structure of neural networks. In this case, it is called the first level of the network, - the second level of the network, and so on. The total length of the chain gives an idea of the model as a whole. It is from this terminology that the name "deep learning" arises.

Based on this principle of neural networks, their ability to detect hidden structures inside unstructured data, which make up the vast majority of data in the world, follows. Raw data in the field of broadcasting can include raw media – images, texts, video and audio recordings. Therefore, one of the problems that deep learning solves best is the processing and clustering of raw, unmarked media, the detection of similarities and anomalies in data that a person is not able to process.

For example, a deep learning neural network can analyze a million images and group them according to similarity: animals in one group, sports in another group, and personal photos of the user in the third group.

A similar idea applies to all other data types. This implies the assumption that a neural network can group voice commands that are given as part of the production of television programs and, by automating the process, ensure the operation of the system without operator actions.

Using the example of an elementary scheme, you can imagine how a neural network for recognizing an audio file will work. Assume that there is an audio file in a format suitable for processing, and upload it to a deep learning neural network. The input data for the neural network will be 30 millisecond audio fragments. For each fragment of sound, the neural network will have to determine the letter corresponding to the sound pronounced at a particular time. In order for the network to be able to recognize voice and sounds received from other sources in the future, we use a recurrent neural network, since such a network has memory and is able to influence the prediction of processes (Figure 1).

Figure 1 – Elementary scheme of training a neural network to recognize soundsIn the diagram shown in Figure 1, a recurrent neural network with memory will affect the prediction of processes in such a way that each letter that is recognized by the network will affect the "probable" prediction of the next letter by the network.

For example, if three letters "at" are pronounced, then the network can predict the "probable" appearance of the following letters "vet" in order to get the finished word "hello". Much less likely for the network will be just a set of letters "adt".After the entire audio file is passed through the neural network, the output will be a comparison of each audio file with the letters that are most likely pronounced during a particular fragment.

In the theoretical literature on the work of neural networks, it is noted that in the process of deep learning, neural networks learn to recognize correlations between certain relevant features and optimal results – they draw connections between feature signals and what these features represent ^[8].

At the same time, a deep learning network built on certain structured data can then be applied to unstructured data, giving them access to a much larger amount of input data than machine learning networks. Thus, the more data the network can train, the more accurate the result will be.

Adapting the theory of neural network training to the solution of a specific task of automation of broadcast production complexes, it can be stated that when developing an appropriate neural network training model and choosing an effective learning mechanism, it is possible to simplify individual processes of program production, for example, by replacing the physical actions of an operator or director with voice commands. At the same time, this assumption requires additional study and the development of an appropriate model adapted to a specific task.

Conclusions

Summing up this research, it should be noted that the existing problems in the field of automation of broadcasting and production complexes at the moment are due to the desire of TV and radio companies to replace the traditional approach to the production of programs using a team of specialists with software. However, such an approach will be effective only if the release of programs is not related to live work, where in an uncertain situation it is not possible to react promptly to any events by introducing a command into the software. Hence, automation of production processes becomes an effective and economical technology only with the correct configuration of the software and careful calculation of automation prospects.

The real prospects for automation of broadcast production complexes are currently in practically unexplored areas for the television sphere - the areas of voice control, which is based on artificial intelligence based on various algorithms for training artificial neural networks, which requires additional study and development of an appropriate model adapted to a specific task.