Method for changing at least one lighting state by voice command, computer system and computer program product for the execution thereof

Abstract

The invention relates to a method for changing at least one lighting state by means of a voice instruction, comprising the steps: obtaining a voice instruction and inputting into a language model; interpreting the voice instruction and breaking down into at least one subtask; acquiring at least one lighting feature of at least one lighting object and at least one setting feasibility of the lighting feature; determining at least one lighting object and at least one lighting feature of the determined lighting object by means of an inference through a trained machine learning model with input of the lighting feature, the setting feasibility and the subtask; creating a schedule by means of an inference through a trained machine learning model with input of the determined lighting object, the determined lighting feature and / or the setting feasibility; and changing the lighting state according to the schedule. The invention also relates to a computer system configured to perform the method and a computer program product comprising software code sections which are configured in such a way that the method can be executed by at least one processor.

Description

Technical Field

[0001] The present invention relates to a method for changing at least one lighting state by means of a voice command, a computer system configured to perform the method, and a computer program product including a software code segment by means of the software code segment for performing the method for changing at least one lighting state by means of a voice command. Background Technology

[0002] It is known in the prior art that light shows are used to highlight performances, such as concerts or stage plays, as well as purely auditory performances, such as musical works in nightclubs, and / or to enhance the visibility of displays. With advancements in lighting technology, more and more possibilities are available for operators. Therefore, sophisticated control technologies are needed to automatically manipulate lighting systems not only in terms of output light but also in terms of motion. Different control protocols are implemented to operate individual lighting mechanisms, which must be programmed with partially different instructions. Light shows are typically pre-programmed, either partially or completely, and played during the performance, allowing for adjustments within a limited range.

[0003] The extensive use of programming in light shows, as mentioned above, is complex and can only be performed by trained personnel. On the one hand, these personnel must master programming languages ​​developed in part by control technology manufacturers, resulting in a lack of industry standardization. On the other hand, they must have an accurate understanding of the existing lighting equipment layout and configuration, as well as the site conditions, in order to create an attractive and coordinated light show.

[0004] In the prior art, it is rare to find assistance systems that suggest ideas to a user based on pre-set parameters, a library of light shows, or previous light shows. Furthermore, a method is known from US 11,687,760 B2 that, after inputting a first instruction, suggests at least one additional instruction to the user using multiple pre-created light shows, which statistically often follow the instruction input up to that point.

[0005] The creation, modification, and adjustment of light shows remain complex processes that can only be implemented after thorough training. Applications are becoming increasingly complex due to advancements in technological feasibility. Summary of the Invention

[0006] Therefore, there is a great need for a method that can change the lighting state of objects used in a light show in a simple and intuitive way, and that can be used in all conceivable applications. Another concern is that the method can be used with all devices known in the prior art and implemented cost-effectively. In this manner, the creation and / or adjustment of light shows can be largely supported and / or automated. Therefore, the object of the present invention is to provide a method, a computer system, and a computer program product to overcome the aforementioned difficulties.

[0007] The objective is achieved in a surprisingly simple yet effective manner by a method according to the teachings of the present invention, a computer system according to the teachings of the present invention, and a computer program product according to the teachings of the present invention.

[0008] According to the present invention, a method for changing at least one lighting state via voice command is provided, wherein the method includes the following steps:

[0009] a. Obtain the voice command in the first step and input the voice command into the language model;

[0010] b. In the second step, interpret the voice instructions and break them down into at least one sub-task;

[0011] c. In the third step, obtain at least one lighting feature of at least one lighting object and at least one setting feasibility of at least one lighting state, and determine at least one lighting object and at least one lighting feature of the determined at least one lighting object by means of inference through a trained machine learning model, given at least one lighting feature, at least one setting feasibility and at least one subtask.

[0012] d. In the fourth step, given the input of at least one identified lighting object, at least one identified lighting feature, and / or at least one setting feasibility, a schedule is created by inference through a trained machine learning model; and

[0013] e. In step five, change at least one lighting state according to the schedule.

[0014] This invention is based on the fundamental concept that the continuous development of language models has led to the use of these models to interpret user instructions written in natural language and ultimately translate them into changes to one or more lighting objects for use in light shows and / or the creation of light shows. This requires, firstly, breaking down voice instructions into sub-tasks and then using a machine learning model specifically trained for this purpose to translate them into corresponding control codes and / or other possible maneuverability and / or visualization for the user to implement the sub-tasks.

[0015] In the first step a., the voice command is obtained and input into the language model. This enables the implementation of the subsequent step b. Preferably, the command is obtained via a corresponding interface through real-time input from a user applying the method. Suitable interfaces are mentioned elsewhere. Preferably, the language model used is specifically trained for the application of the method according to the invention. More preferably, the language model is a Large Language Model (LLM), especially a Generative Pre-trained Transformer (GPT).

[0016] In the next step, this second step b, the speech instruction is interpreted by a language model and decomposed into at least one subtask. In other words, the language model breaks down the speech instruction into aspects that must be processed to achieve the desired change. Preferably, the speech instruction is decomposed into at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, or 100 subtasks. These subtasks may, in particular, involve changing different lighting characteristics of the lighting state of a single object, individual lighting characteristics of different lighting states of different objects, different lighting characteristics of different objects, and / or other aspects beyond the lighting object, especially aspects that directly or indirectly affect at least one lighting object. Particularly preferably, at least one subtask is converted into a pre-defined syntax. This is simple to implement due to the limited feasibility of the changes and leads to better and / or more accurate results.

[0017] In the next step, this third step c, at least one lighting feature of at least one lighting object and at least one setting feasibility of at least one lighting object are obtained, wherein at least one lighting feature, at least one setting feasibility, and at least one subtask are input into a trained machine learning model, wherein the trained machine learning model determines at least one lighting object and at least one lighting feature of a previously determined lighting object by inference. Upon determination, at least one lighting object becomes the determined at least one lighting object, and at least one lighting feature becomes the determined at least one lighting feature. In order to induce a desired change in a voice command, in most cases, it is necessary and / or desired to change the individual lighting features and / or the individual lighting objects. Obtaining the lighting features may here involve the type of the lighting feature, the value of the lighting feature, and / or the value range of the lighting feature. In other words, it is conceivable that obtaining the lighting features involves collecting knowledge about whether a determined lighting feature exists in at least one lighting object, the specific setting of the lighting feature, and / or a technically feasible setting. Preferably, technically feasible settings are considered such that, upon determination, at least one lighting object and / or at least one lighting feature is prevented from being determined due to its setting feasibility, thus preventing the determination of at least one lighting object and / or at least one lighting feature from achieving the desired change. The feasibility of obtaining at least one lighting feature and / or at least one setting is preferably achieved by retrieving and / or retrieving stored data from a database, by querying the setting of at least one lighting feature, by evaluating previous control commands, and / or by querying at least one state sensor. This determination can be accomplished by training a trained machine learning model using a training dataset prior to implementing the method. Preferably, the training is supervised learning, unsupervised learning, or reinforcement learning. The determination, in particular, leads to the selection and / or specification of assumed values ​​for the lighting feature and / or the lighting object. Preferably, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 100, 150, 200, 250, 300, 350, 400, 450, 500, 550, 600, 650, 700, 750, 800, 850, 900, 950, 1000, 1500, 2000, 25 00, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 6500, 7000, 7500, 8000, 8500, 9000, 9500, 10000, 15000, 20000, 25000, 30000, 35000, 40000, 45000, 50000, 55000, 60000, 65000, 70000, 75000, 80000, 85000, 90000, 95000, 100000150000, 200000, 250000, 300000, 350000, 400000, 450000, 500000, 550000, 600000, 650000, 700000, 750000, 800000, 850000, 900000, 950000, 1000000, 1500000, 2000000, 2500000, 300000 0, 3500000, 4000000, 4500000, 5000000, 5500000, 6000000, 6500000, 7000000, 7500000, 8000000, 8500000, 9000000, 9500000, 10000000, 15000000, 20000000, 25000000, 30000000, 350000 00, 40000000, 45000000, 50000000, 55000000, 60000000, 65000000, 70000000, 75000000, 80000000, 85000000, 90000000, 95000000, 100000000, 150000000, 200000000, 250000000, 300000 000, 350000000, 400000000, 450000000, 500000000, 550000000, 600000000, 650000000, 700000000, 750000000, 800000000, 850000000, 900000000, 950000000, or 1000000000 different or identical lighting features. More preferably, the lighting state of at least one lighting object is acquired. More preferably, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 100, 150, 200, 250, 300, 350, 400, 450, 500, 550, 600, 650, 700, 750, 800, 850, 900, 950 are obtained. 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 6500, 7000, 7500, 8000, 8500, 9000, 9500, 10000, 15000, 20000, 25000, 30000, 35000, 40000, 45000, 50000, 55000, 60000, 65000, 70000, 7500080000, 85000, 90000, 95000, 100000, 150000, 200000, 250000, 300000, 350000, 400000, 450000, 500000, 550000, 600000, 650000, 700000, 750000, 800000, 850000, 900000, 950000, 1000000, 1500000, 200000 0, 2500000, 3000000, 3500000, 4000000, 4500000, 5000000, 5500000, 6000000, 6500000, 7000000, 7500000, 8000000, 8500000, 9000000, 9500000, 10000000, 15000000, 20000000, 25000000, 30000000, 35000 000, 40000000, 45000000, 50000000, 55000000, 60000000, 65000000, 70000000, 75000000, 80000000, 85000000, 90000000, 95000000, 100000000, 150000000, 200000000, 250000000, 300000000, 350000000, At least one lighting feature of 400,000,000, 450,000,000, 500,000,000, 550,000,000, 600,000,000, 650,000,000, 700,000,000, 750,000,000, 800,000,000, 850,000,000, 900,000,000, 950,000,000, or 100,000,000,000 different or identical lighting objects, which may involve the same and / or different lighting features. Most preferably, the lighting state of all lighting objects is acquired. Thus, particularly precise and comprehensive adjustments are possible; however, depending on the setup, the amount of data acquired is very large. Preferably, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 100, 150, 200, 250, 300, 350, 400, 4 50, 500, 550, 600, 650, 700, 750, 800, 850, 900, 950, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 6500, 7000, 7500, 8000, 8500, 90009500, 10000, 15000, 20000, 25000, 30000, 35000, 40000, 45000, 50000, 55000, 60000, 65000, 70000, 75000, 80000, 85000, 90000, 95000, 100000, 150000, 200000, 250000, 300000, 350000, 400000, 450000, 500000, 550000, 60 0000, 650000, 700000, 750000, 800000, 850000, 900000, 950000, 1000000, 1500000, 2000000, 2500000, 3000000, 3500000, 4000000, 4500000, 5000000, 5500000, 6000000, 6500000, 7000000, 7500000, 8000000, 8500000, 9000 000, 9500000, 10000000, 15000000, 20000000, 25000000, 30000000, 35000000, 40000000, 45000000, 50000000, 55000000, 60000000, 65000000, 70000000, 75000000, 80000000, 85000000, 90000000, 95000000, 100000000, 1 Feasibility of setting 50000000, 200000000, 250000000, 300000000, 350000000, 400000000, 450000000, 50000000, 550000000, 600000000, 650000000, 700000000, 750000000, 800000000, 850000000, 900000000, 950000000, or 1000000000. More preferably, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 100, 150, 200, 250, 300, 350, 400 are obtained. 450, 500, 550, 600, 650, 700, 750, 800, 850, 900, 950, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 6500, 7000, 7500, 8000, 8500, 90009500, 10000, 15000, 20000, 25000, 30000, 35000, 40000, 45000, 50000, 55000, 60000, 65000, 70000, 75000, 80000, 85000, 90000, 95000, 100000, 150000, 200000, 250000, 300000, 350000, 400000, 450000, 500000, 550000, 600000, 6500 00, 700000, 750000, 800000, 850000, 900000, 950000, 1000000, 1500000, 2000000, 2500000, 3000000, 3500000, 4000000, 4500000, 5000000, 5500000, 6000000, 6500000, 7000000, 7500000, 8000000, 8500000, 9000000, 9500000, 100000 00, 15000000, 20000000, 25000000, 30000000, 35000000, 40000000, 45000000, 50000000, 55000000, 60000000, 65000000, 70000000, 75000000, 80000000, 85000000, 90000000, 95000000, 100000000, 150000000, 200000000, 250000000 The feasibility of setting at least one of the following: 300,000,000, 350,000,000, 400,000,000, 450,000,000, 500,000,000, 550,000,000, 600,000,000, 650,000,000, 700,000,000, 750,000,000, 800,000,000, 850,000,000, 900,000,000, 950,000,000, or 100,000,000, where the feasibility of the same and / or different settings may be involved. Most preferably, all feasibility settings for all lighting objects are obtained. This allows for particularly precise and comprehensive adjustments; however, depending on the setting, the amount of data obtained is very large. When determining, it is preferable to determine a number of lighting objects and / or lighting features that is less than the total number of available lighting objects. More preferably, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 100, 150, 200, 250, 300, 350 are determined.400, 450, 500, 550, 600, 650, 700, 750, 800, 850, 900, 950, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 6500, 7000, 7500, 8000, 8500, 9000, 9500, 10000, 15000, 20000, 25000, 30000, 35000, 40000, 45000, 50000, 55000, 60000, 65000, 70000, 7500 0, 80000, 85000, 90000, 95000, 100000, 150000, 200000, 250000, 300000, 350000, 400000, 450000, 500000, 550000, 600000, 650000, 700000, 750000, 800000, 850000, 900000, 950000, 1000000, 1500000, 2000000, 2500000, 3000000, 3500000, 4000000, 4500000, 5000 000, 5500000, 6000000, 6500000, 7000000, 7500000, 8000000, 8500000, 9000000, 9500000, 10000000, 15000000, 20000000, 25000000, 30000000, 35000000, 40000000, 45000000, 50000000, 55000000, 60000000, 65000000, 70000000, 75000000, 80000000, 85000000 90,000,000, 95,000,000, 10,000,000, 15,000,000, 20,000,000, 25,000,000, 30,000,000, 35,000,000, 40,000,000, 45,000,000, 50,000,000, 55,000,000, 60,000,000, 65,000,000, 70,000,000, 75,000,000, 80,000,000, 85,000,000, 90,000,000, 95,000,000, or 100,000,000,000 lighting objects and / or lighting features. More preferably, at least one lighting characteristic and / or at least one setting feasibility of the identified at least one lighting object is determined, wherein the lighting object should undergo change or the lighting object should cause such change. Most preferably, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, of the identified at least one lighting object are determined.Feasibility of setting 15, 16, 17, 18, 19, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 100, 150, 200, 250, 300, 350, 400, 450, 500, 550, 600, 650, 700, 750, 800, 850, 900, 950, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 6500, 7000, 7500, 8000, 8500, 9000, 9500, or 10000.

[0018] In the next step, this fourth step d, a schedule is created by inference by inputting the identified at least one lighting object, the identified at least one lighting feature, and / or at least one setting feasibility into a trained machine learning model. Creating the schedule by inference is feasible by training the trained machine learning model with a training dataset prior to implementing the method. Preferably, the training is supervised learning, unsupervised learning, or reinforcement learning. The schedule implements all sub-tasks according to the user's presets. More preferably, during and / or after implementing the method according to the invention, the trained machine learning model is updated by the user evaluating the created schedule, or by the user evaluating intermediate information fed back during the implementation of the method, and the evaluation is fed into the machine learning model as a new training dataset. This allows for continuous improvement and personalization of the trained machine learning model by the user. More preferably, in steps c and / or d, at least one setting feasibility for setting the identified at least one lighting object and / or at least one identified lighting feature is determined.

[0019] The trained machine learning model is preferably two-part, with the first part used in step c. and the second part used in step d. This allows for easy reading of intermediate results, particularly results identifying at least one illuminated object and / or at least one illumination feature, thereby facilitating error correction through adjustments. It is also conceivable that two independent trained machine learning models are involved, where the trained machine learning model in step c. is a first trained machine learning model, and the trained machine learning model in step d. is a second trained machine learning model. Even more preferably, the trained machine learning model is one-part, thereby allowing steps c. and d. to be implemented and / or repeated at least once, either partially or completely in parallel. This results in better results with faster implementation times. Particularly preferred are the trained machine learning models, especially recurrent neural networks (RNNs), feedforward neural networks (FNNs), convolutional neural networks (CNNs), transformers, flow-based generative models, evolutionary neural networks, encoder-decoder models, variational autoencoders, autoregressive models (ARMA models), restricted Boltzmann machines (RBMs) and / or diffusion models, hidden Markov models (HMMs) and / or support vector machines (SVMs).Furthermore, it is conceivable to use genetic programming, boosting, decision tree machine learning, kernel density estimation (KDE), expert systems (ES), (naive) Bayesian classifiers, gradient boosting, linear discriminant analysis, nearest neighbor classification, clustering methods, especially single-linkage methods, complete-linkage methods, Ward methods, K-means algorithm, fuzzy C-means algorithm, expectation-maximization algorithm (EM algorithm), DBSAN (Density-Based Spatial Clustering of Applications with Noise), STING algorithm (Statistical Information Grid-based Clustering Algorithm), and / or CLI-QUE algorithm (Clustering Query Algorithm). Inquest-Algorithms and / or anomaly detection methods, particularly Local Outlier Factor (LOF), Isolation Forest, and / or autoencoders and / or Principal Component Analysis (PCA). Furthermore, reinforcement learning methods, such as Associative Reinforced Learning, Deep Reinforcement Learning, Adversarial Deep Reinforcement Learning, Fuzzy Reinforcement Learning, and / or Safe Reinforcement Learning, can be used. In particular, it is conceivable that methods for clustering the data can also be used. Suitable measures for creating, using, and / or training are known to those skilled in the art. It is also conceivable that training data be stored in a database, which is continuously expanded with new training data during operation or when using the method according to the invention. It is also conceivable that the trained machine learning model is the language model from step b.Other machine learning models and the feasibility of creating, using, and / or training said machine learning models are known to those skilled in the art.

[0020] It is also conceivable that after implementing steps a., b., c., and / or d., the agent system checks and verifies the results, namely, the determination of at least one subtask, at least one lighting object, and / or the determination of at least one lighting characteristic of the determined at least one lighting object and / or the schedule. It is conceivable that if the check result is negative, steps a., b., c., and / or d. are repeated partially or entirely with modifications to the input. In a preferred extension of the agent system, multiple agents with dedicated responsibilities are used, such as those specifically for checking results, analyzing error messages, general monitoring, and dividing subtasks and / or implementing the instructions themselves.

[0021] Preferably, the schedule includes at least one simulation dataset, at least one model, especially a virtual 3D model, at least one cue, especially a series of cuees, at least one preset, at least one sequence, at least one stack, at least one video file, especially video recordings and / or simulations, and / or at least one data packet, especially including recipes, MAtricks, phasers, timecodes, macros, Lua plugins, filters, selections, effects, bitmaps, and / or generators. Particularly preferably, the schedule includes DMX data packets or Artnet data packets. This relates to the most widely used protocols for controlling lighting technology. Particularly preferably, the data packet itself includes at least one preset, at least one cue, at least one sequence, and / or at least one stack. The simulation dataset is preferably a simulation of the stage together with the lighting objects, especially lighting fixtures, set thereon. Further preferably, the model also shows the stage with lighting objects, especially lighting fixtures. Further preferably, the video file, especially video recordings, and / or simulations show the progress of the light show.

[0022] Subsequently, in the next step, this fifth step e., at least one lighting state is changed according to the schedule. It is also conceivable that the schedule is saved in a series of schedules, particularly at the end of the series, to create and save the light show. It is also conceivable that the schedule replaces another schedule in the series. Thus, an existing light show can be modified according to the user's presets. Particularly preferably, changes according to the schedule include movement, especially movement speed, and / or state switching, such as on and off, and changes in the intensity, focal length, shape, and / or color of at least one illuminated object, especially at least one identified illuminated object. Even more preferably, the schedule includes data packets that can be directly transmitted to the lighting control device, especially the light control system, via an output interface.

[0023] The method can be performed based on real lighting objects, especially distant areas, particularly stage lighting installations, or virtually in the form of simulation. Furthermore, the method can be performed based on virtual lighting objects, particularly parameters in control code. Moreover, the method can be partially or fully automated. Commonly in light show technology, a complex light show is first created in a simulation program, and then the light show is transmitted to a real stage technology mapped through the simulation. The schedule can also take into account possible limitations due to the expectations of the lighting objects or users. In particular, it is conceivable that the schedule first proposes: turning off the light-emitting mechanism of the lighting object before it moves to its final position. Upon reaching the final position, the light-emitting mechanism of the lighting object is reactivated, and the lighting characteristics proposed according to the schedule are set.

[0024] As will be understood by those skilled in the art, the previously described method includes at least the five steps mentioned, which may be repeated individually or collectively with arbitrary frequency. It is conceivable that the method may also include other steps as described elsewhere, and / or the described steps may be decomposed into partial steps, sub-steps, or sub-tasks.

[0025] The method according to the invention allows users to express their desire for change through voice commands in a simple yet effective manner, without the need for complex programming languages. In particular, users can use everyday language and / or spoken language. This enables the simple and rapid creation of light shows without prior training.

[0026] The term "language model" refers to a mathematical model that models the order of elements in a sequence of natural language and is adapted to detect and separate at least one subtask in natural language.

[0027] The term "subtask" refers to a user-expressed expectation of a change in the lighting state of at least one lighting feature, a user-preset selection of at least one lighting object or at least a group of lighting objects whose individual or multiple lighting states should undergo a change, and / or its direct or indirect aspects, and / or the direct or indirect aspects of an existing light show for at least one lighting object.

[0028] The term "lighting object" refers to an object in a light control system having or not having a direct physical counterpart, wherein changes to the lighting characteristics can be contained in a human-perceptible static or dynamic alteration within the physical expression of the light control system. Lighting objects are, in particular, but by no means limited to, lighting fixtures, speedmasters, cues, presets, sequences, fixtures, temporal sequences, and / or lighting atmospheres.

[0029] The term "lighting device" refers to a device that can produce light for a light show to illuminate and / or for effect purposes, and in particular can be controlled by a light control system.

[0030] The term "cue" refers to a container of data, particularly a illuminated object, corresponding to a single lighting atmosphere or appearance of a stage performance, wherein the cue is typically pre-set and / or pre-settable.

[0031] The term "preset" refers to a container of a limited number of expressions of data, particularly of a lighting object, used to create a specific lighting feature for a cue.

[0032] The term "sequence" refers to a list and / or arrangement of prompts.

[0033] The term "type of lighting device" refers to the variety of lighting devices. Specifically, the type of lighting device can be a PAR spotlight, flashlight, floodlight, lens spotlight, moving head light, scanner, laser show, LED spotlight, surface light, horizontal light, and / or moving light. Other lighting devices and methods for operating said lighting devices are known to those skilled in the art.

[0034] The term "lighting characteristic" refers to a single feature of a lit object or a value representing that single feature and / or the defined range of influence of that value. Lighting characteristics include, but are not limited to, state, color, hue, color temperature, brightness, zoom, focal length, aperture, shape, orientation, rotation, position, and / or speed.

[0035] The term "lighting state" refers to the set of all lighting characteristics of a lit object.

[0036] The term "state of a light control system" refers to the set of lighting states of all illuminated objects in a light control system.

[0037] The term “acquisition” refers to the collection of information and / or knowledge that can be provided to a machine learning model in a processable manner.

[0038] The term "machine learning model" refers to a program configured to identify statistical relationships, patterns, and / or structures among information contained in a training dataset without explicit pre-defined programming, and to obtain at least one output based on the information obtained as input.

[0039] The term "inference" refers to deriving at least one output using a machine learning model created from a training dataset.

[0040] The term "determine" specifically relates to: the selection of at least one lighting object, wherein at least one lighting feature of the determined at least one lighting object shall be changed according to a user preset; the selection of at least one lighting feature, wherein at least one lighting feature shall be changed according to a user preset; and / or the specification of at least one value of the determined at least one lighting feature, the value of which shall correspond to a user preset.

[0041] The term "agent" refers to a system that acts autonomously by making decisions based on input information.

[0042] Advantageous improvements to the invention are shown below, which can be implemented individually or in combination.

[0043] It is conceivable that the method includes a next step, namely step f, after step d and / or after step e, wherein step f includes the following:

[0044] f. Obtain at least one altered lighting feature and verify the at least one altered lighting feature.

[0045] During verification, it is particularly preferred to check whether the lighting features have been changed in accordance with the schedule and the sub-task presets. Verification can also check whether the implementation of steps b. to e. has resulted in the expected results. Particularly preferred, verification is performed by an agent. More preferably, the agent is a machine learning model trained for verification or a deterministic algorithm oriented towards verification. For verification, in addition to the at least one changed lighting feature, it is preferred to provide the agent with at least one sub-task, namely, the determination of at least one lighting feature, the determination of at least one lighting object, and / or the determination of at least one lighting feature, and / or the schedule. The at least one lighting feature here relates to the setting of the changed lighting feature before the change. Providing more information makes the verification more reliable. Even more preferably, all lighting features changed in step e. are verified in step f. Most preferably, the state of the lighting system is acquired and verified. Through complete detection, lighting features that have not been changed but should undergo changes according to the user's voice commands can be detected and considered during verification. Additionally or alternatively, verification can be performed by querying and / or asking the user.

[0046] In one improved embodiment, step f may further include: if the verification yields a negative result, modifying the subtask, determining and / or scheduling, and repeating steps b, c, d, e, and / or f. If the changed lighting feature and / or the change in the lighting feature does not conform to the preset of the subtask and / or schedule, a corresponding new schedule is created through modification to achieve the desired objective. It is evident here that the steps involving modification are repeated depending on the location of the deviation or the location where modification is made. If the schedule is modified, it is conceivable not to re-implement or only partially re-implement steps b and c. However, it is conceivable to re-implement step c to determine at least one other lighting object and / or at least one other lighting feature if necessary. Preferably, it is conceivable that the modification includes reverting at least one changed lighting feature.

[0047] It is also conceivable that the voice instructions in step a. exist orally and / or in writing. Oral voice instructions can be input via an interface designed as a microphone. Written voice instructions can be input, in particular, via an input interface, especially a keyboard. It is conceivable that the voice instructions include not only spoken portions but also written portions, which are combined together. It is also conceivable that the language model in step a. converts the spoken portions into written portions, especially text. Furthermore, it is conceivable that the spoken voice instructions are recorded elsewhere and played back on-site.

[0048] The term "oral" refers to spoken language.

[0049] The term "written" refers to written language.

[0050] Furthermore, it is conceivable that the schedule includes implementable control code for the light control system, particularly the light control console, and that the changes in step e are made by implementing the control code. Thus, the lighting technology at the activity location can be changed directly and indirectly via voice commands using the method according to the invention. The control code preferably includes at least one control command.

[0051] The term "light control system" refers to software, particularly software installed on a light control console, having an interface for controlling at least one real lighting device and an interface to a human user, which allows the user to control the shape, focal length, color, intensity, and / or orientation of the light emitted by the lighting device, where technically feasible and supported.

[0052] In an improved version of the method, it is conceivable that in step d., a schedule is created by inference from a trained machine learning model given the input subtask. Thus, the subtask can be reconsidered at this stage of the method, thereby achieving better and more accurate results. Intermediate validation can also be performed via a proxy or proxy system to prevent erroneous schedules. In addition to the subtask, at least one identified lighting object, at least one identified lighting feature, and / or at least one setting feasibility are also input into the trained machine learning model.

[0053] The term “consider” refers to providing information to be considered in a detectable manner, such that trained machine learning and / or language models can, however, not necessarily, identify statistical relationships, patterns, and / or structures, particularly with other inputs and / or information.

[0054] Furthermore, it is conceivable that in step b., at least one subtask is created, taking into account the context. This allows for the incorporation of additional information when processing voice commands, thereby enabling users to use language more naturally. This prevents erroneous implementation of the schedule and achieves more accurate results. Additionally, it is possible to extract information from the dialogue of at least two users. Particularly preferably, the context consists of at least one earlier voice command that the user has already entered before the current voice command. Therefore, the user does not have to express a complete, self-understanding voice command every time. Preferably, at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 earlier voice commands are considered, especially all earlier voice commands.

[0055] Furthermore, it is conceivable that in step b., at least one subtask is created; in step c., at least one lighting object and at least one lighting feature of the determined lighting object are determined by inference through a trained machine learning model; and / or in step d., a schedule is created by inference through a trained machine learning model, taking into account at least one additional factor. Preferably, the additional factor is a musical composition, location features, the additional status of at least one additional device, a light show, and / or a sketch. Considering the musical composition allows for a light show that is better coordinated with the planned performance, which simplifies the creation of the light show for the user. Location features are particularly those related to the event location, especially whether it involves a hall or an open-air venue. In addition, time features can be considered, from which the ambient brightness can be inferred. Additional devices can be devices commonly used in stage technology, such as curtains, screens, and / or effects equipment. It is also conceivable here to create video sequences for the screen by means of an agent. In particular, inputting existing light shows as control codes can enable adjustments to the light show for the situation and / or according to the user's individualized expectations. The sketch can particularly include stage design and / or stage construction.

[0056] Furthermore, it is conceivable that in step c., at least one signal from at least one sensor and / or at least one transmitter is acquired, wherein in step c., at least one lighting object and at least one lighting feature are determined by inference through a trained machine learning model upon input of at least one signal, and / or in step d., a schedule is created by inference through a trained machine learning model upon input of at least one signal. External, however important data for the light show can be considered by means of the sensors and / or transmitters. The sensors and / or transmitters can detect information about external data, such as ambient brightness, ambient background sound effects, and / or events occurring on stage. In particular, the movement of people acting on stage can be detected by means of sensors. Particularly preferably, the sensor is a camera that detects the stage. More preferably, the transmitter is part of a tracker. Thus, stage performances can be better coordinated with events occurring on stage by means of voice commands. In addition to signals, other factors mentioned elsewhere can or must be input and / or considered in steps c. and / or d.

[0057] In another embodiment of the invention, it is conceivable that the voice command includes control codes for at least one first naming scheme, wherein steps b. and / or c. include converting the control codes for at least one first naming scheme into a second naming scheme. It has been recognized within the scope of this invention that a common problem in lighting technology control involves different naming schemes from different manufacturers. Users are typically trained particularly well with at least one established first naming scheme, thus enabling them to use the first naming scheme particularly well. If a second naming scheme is used in the setting to which the method is applied, it is advantageous that the language model is trained to convert at least one first naming scheme into the second naming scheme.

[0058] Furthermore, it is conceivable that voice commands are given in one or more arbitrary languages, with a language model translating the voice commands into a unified language. Therefore, it is irrelevant which language the user speaks.

[0059] It is also conceivable that the method in step b. includes: evaluating the voice instruction and / or at least one subtask, and requesting additional voice instruction if the verification yields a negative result. Thus, in an early stage, the voice instruction can be interpreted as follows: whether all necessary information for implementing the voice instruction is known and / or explicit. In cases of ambiguity and / or missing information, the information can be simply supplemented by the user. Particularly advantageously, consider the first voice instruction and the voice instruction given in response to the request. In other words, the first voice instruction is the context considered by the language model, described elsewhere as an earlier voice instruction. The request is preferably given in written and / or oral form.

[0060] Furthermore, it is conceivable to obtain at least one setting feasibility for at least one lighting state by accessing a database. The database stores various information about at least one lighting object, particularly at least one lighting fixture, including, in particular, the type of lighting fixture, the setting feasibility for the lighting object, the manner in which it can be set, the location of the lighting fixture, possible grouping with other lighting objects, and / or the possible orientation of the lighting fixture. In this manner, the type of setting feasibility and its scope can be obtained quickly and easily.

[0061] Unless otherwise stated, it is assumed that the definitions and / or embodiments of the terms mentioned above apply to all aspects described below. In particular, the aspects of the computer system or computer program product disclosed in the methods described above are possible implementations of the computer system or computer program product described below, and the aspects of the computer system or computer program product disclosed below that are features of the method are conceivable features of the method described above.

[0062] According to the present invention, a computer system is also provided, wherein the computer system is configured to implement at least one of the methods described above, and the computer system includes at least one interface for obtaining voice instructions to implement step a., a computer-readable storage medium for storing the voice instructions and a trained machine learning model to implement steps b. to d., and at least one data processing means configured to implement steps b. and d.

[0063] The advantages disclosed in the method can be achieved by means of the computer system. Preferably, the computer system includes at least one sensor interface and / or transmitter interface for receiving signals emitted by sensors and / or transmitters. More preferably, the interface is a microphone and / or keyboard. Furthermore, the computer system preferably includes an output interface by means of which schedules and / or control codes can be transmitted to the optical control system. More preferably, the computer system includes an output interface by means of which language, especially written or spoken language, can be output. Particularly preferably, the output interface includes a screen and / or speakers. The computer system can be a physical computer system in the field, a cloud-based computer system connected via a remote communication interface, and / or a hybrid thereof.

[0064] Furthermore, according to the present invention, a computer program product is provided, the computer program product comprising software code segments configured such that one of the methods described above can be implemented by means of at least one processor. Attached Figure Description

[0065] Further details, features, and advantages of the invention will become apparent from the following description of preferred embodiments in conjunction with the specification. Here, the corresponding features may be implemented individually or in combination with each other. The invention is not limited to the embodiments. Embodiments are schematically illustrated in the accompanying drawings. The same reference numerals in the various figures denote the same or functionally identical elements or elements that correspond to each other in function.

[0066] Specifically shown:

[0067] Figure 1 A progress chart of the method according to the invention is shown; and

[0068] Figure 2 A computer system according to the present invention is shown. Detailed Implementation

[0069] Figure 1 A progress diagram of the method according to the invention is shown. User 1 speaks voice command 2, which is transmitted through a suitable interface 21 (see...). Figure 2For example, a microphone is used to record. In step a.101, the voice command 2 is fed to the language model 3. Then, the first sub-step 102 of step b. is performed, in which the voice command 3 is decomposed into three sub-tasks 4 by the language model 3, and in the subsequent second sub-step 103 of step b., these three sub-tasks 4 are arranged sequentially. In the subsequent first sub-step 104 of step c., at least one lighting feature 6 of the three lighting objects 7 and its setting feasibility 8 are obtained, for example, by reading data from a database and / or stored in the light control system. In addition to the sub-tasks 4, the obtained lighting feature 6 and the obtained setting feasibility 8 are also input into a trained machine learning model 5, which evaluates and weights them in the subsequent second sub-step 105 of step c. Then, the third sub-step 106 of step c. is performed, in which the trained machine learning model 5 determines, by inference, one of the lighting objects 7 and at least one of the lighting features 6 of the determined lighting object 7 based on the sub-task 4, the lighting features 6 and the setting feasibility 8, wherein it is now the determined lighting object 7 and the determined lighting feature 6.

[0070] In the subsequent first sub-step 107 of step d., the feasibility 8 of setting the identified lighting object 7 is determined by a trained machine learning model 5 in such a way that the trained machine learning model 5 determines and / or creates control instructions 9 by means of inference. Subsequently, in the second sub-step 108 of step d., the obtained control instructions 9 are arranged in the correct order and the instructions are converted into a schedule 10, which is designed as control code.

[0071] Subsequently, in the first sub-step 109 of step f., it is verified whether schedule 10 is a schedule 10 that can be implemented on the light control system. If not, the (sub)step is reimplemented starting from the first sub-step 107 of step d. If the expected effect achieved by control command 9 has been achieved, schedule 10 is implemented in step e.110, and compared with subtask 4 and / or voice command 2 expressed by user 1 in the second sub-step 111 of step f. Here, user 1 is requested to provide feedback. If the result does not conform to subtask 4, the subtask is modified, and the (sub)step is reimplemented starting from the first sub-step 102 of step c. or the second sub-step 105 of step c. depending on the deviation location. If the result conforms to the target defined in subtask 4 and / or voice command 2 expressed by user 1, the method ends in the final step 112 with schedule 10 stored.

[0072] Figure 2A computer system 20 according to the present invention is shown. The computer system 20 includes an interface 21 for obtaining voice commands, a computer-readable storage medium 22 including a trained machine learning model 5 and a language model 3, and a data processing device 23. Furthermore, the computer system 20 includes output interfaces 24 and 25 for transmitting a schedule 10 and / or control codes to an optical control console, by means of which feedback as language output can be output to the user.

[0073] The following examples are for illustrative purposes only. These examples should not be construed as limiting the scope of the invention in any way.

[0074] Example 1

[0075] The following is based on Figure 1 A first example of the method according to the present invention is presented.

[0076] The voice command 2 given by user 1 is "Set the back of the stage to Congo blue".

[0077] In the first sub-step 102 of step b, subtask 4 is interpreted from voice instruction 2. Part of subtask 4 involves determining the lighting object 7 designed as a lighting device from the “back of the stage” part of voice instruction 2. The trained machine learning model 5 here can determine this, in particular, in the third sub-step 106 of step c, based on the knowledge of the position of the lighting object 7 that can illuminate the back of the stage, and / or based on the knowledge of the degrees of freedom of the lighting object 7 that can realize the lighting of the back of the stage, through statistical relationships about the group names of the lighting object 7, for example, grouped by the group name “Backtruss”.

[0078] A matching lighting feature 6 must be determined for the illuminated object 7. This is another part of subtask 4 derived from the "Congo Blue" part of voice instruction 2. To do this, the technical expression "Congo Blue" must first be interpreted, which can be undertaken by language model 3 in step b.102 or by trained machine learning model 5, particularly in the first sub-step 104 of step c. This can be achieved, in particular, by combining color presets from databases, such as gel pools, farb-preset pools, and / or user-created color presets, and / or by previously learned knowledge.

[0079] The trained machine learning model 5, in subsequent steps, namely the second sub-step 105 of step c. to the second sub-step 108 of step d., creates a schedule 10 taking into account the feasibility 8 of setting lighting features 6, such as, in particular, the color system used by the lighting object 7 (RGB, HSB, or CIE) and / or orientation settings, such as Pan / Tilt or XYZ coordinates, and as determined. Here, the feasibility 8 of setting brightness, zoom, focal length, aperture, shape, rotation, and / or position should be considered in particular.

[0080] To achieve a consistent final result during implementation in step e., the trained machine learning model 5 should consider the timing of the control commands 9, such that brightness may decrease initially and increase after the color setting, because the illuminated object 7 may not be brightly lit during color switching, potentially causing inconsistent visuals. Brightness should also only be increased when the illuminated object 7 has been moved to a position and / or orientation set according to schedule 10, preventing unwanted light cones from sweeping across the stage. That is, before increasing brightness, the trained machine learning model 5 should automatically pause in accordance with the necessary positional change (Positions-Delta) of the illuminated object 7 and the possible specific duration of the positional change after the set position. In particular, not only at least one lighting feature 6 should be considered, but all earlier voice commands 2 should be considered as context so that the desired intent can be achieved conflict-free and verifiable.

[0081] Example 2

[0082] The voice command 2 given by user 1 is "Please set the lobby to red".

[0083] Basically, the steps are the same as in the first example. It is particularly important to emphasize here that the trained machine learning model 5 and / or language model 3 must also be able to infer, from the data present in the lighting control, the appropriate lighting object 7 designed as a lighting fixture for the lighting of the hall, without explicitly specifying what the “hall” is.

[0084] Here, the trained machine learning model 5 should also infer that, for such a large area of ​​illumination, the light from the illuminated object 7 should be less focused, if feasible. This is achieved by adjusting the zoom and / or focal length to adjust the beam size, which necessitates setting additional illumination features 6 while taking into account the necessary sequence of control instructions 9. At this point, it is advantageous to partially or completely repeat steps c. and d.

[0085] Example 3

[0086] The voice command 2 given by user 1 is "Make the movement faster".

[0087] Basically, the steps are the same as in the first and second examples. However, to achieve the goal, the abstract lighting feature 6, such as the speed of a moving lighting device, must be changed to create the schedule 10. This speed lighting feature 6 typically does not have a direct physical counterpart (Entsprechung) as a directly controllable lighting feature 6 of a real lighting device. Therefore, the speed abstract lighting feature 6 must be changed in the existing control code and / or the previously existing schedule 10.

[0088] Depending on the light control system and its state, this can be done in different ways. In particular, a schedule 10 consisting of control instructions 9 can be created to increase the speed of the motion effects for all light scenes with motion effects. Additionally and / or alternatively, a so-called Speedmaster should be identified as being assigned to all these motions. In this case, the trained machine learning model 5 should interpret and implement the user input as follows: increase the speed of the Speedmaster.

[0089] Example 4

[0090] The voice command 2 given by user 1 is "Move the storm scene after sunset".

[0091] Essentially, the steps are the same as in the first, second, and third examples. Here, it must be inferred what the input "scene storm and sunset" means. The trained machine learning model 5 should infer from general knowledge about light control and knowledge of the overall state of the light control system that it should move the lighting object 7 of type "cue," since the term "scene" is often used as a synonym for the term "cue." Then, the trained machine learning model 5 should infer that "storm" refers, for example, to cue number five in the current cue list, named "lightning and thunder," while "sunset" refers to cue number ten in the same list, in which most of the lighting fixtures in the performance change from yellow to red and then fade out. The machine learning model 5, having acquired this knowledge from the state of the light control system, determines by inference that it generates an instruction that, as part of a sequence of common lighting objects 7, changes the order of cue number five and cue number ten, placing cue number five after cue number ten.

Claims

1. A method for changing at least one lighting state via voice command (2), the method comprising the steps of: a. (101) Obtain the voice command (2) and input the voice command (2) into the language model (3); b. (102) Interpret the voice instruction (2) and break it down into at least one subtask (4), and optionally (103) classify the at least one subtask (4); c. (104) Obtain at least one lighting feature (6) of at least one lighting object (7) and at least one setting feasibility (8) of the at least one lighting feature (6), and (105, 106) In the case of inputting the at least one lighting feature (6), the at least one setting feasibility (8) and the at least one subtask (4), determine at least one lighting object (7) and at least one lighting feature (6) of the determined at least one lighting object (7) by means of inference through a trained machine learning model (5). d. (107, 108) Given at least one determined lighting object (7), at least one determined lighting feature (6) and / or the feasibility of the at least one setting (8), a schedule (10) is created by means of inference through the trained machine learning model (5). e. (110) Change the at least one lighting state according to the schedule (10).

2. The method according to claim 1, The method includes step f. (109, 111) after step d. (107, 108) and / or step f. (110), wherein step f. (109, 111) includes the following: f. (109, 111) Obtain at least one altered lighting feature (6) of at least one altered lighting state and verify at least one altered lighting feature (6).

3. The method according to claim 2, Step f. (109, 111) further includes the following: If the verification has a negative result, modify the subtask (4), the determination and / or the schedule (10), and repeat steps b. (102, 103), c. (104, 105, 106), d. (107, 108), e. (110) and / or f. (109, 111).

4. The method according to any one of the preceding claims, The voice instruction (2) in step a. (101) exists orally and / or in writing.

5. The method according to any one of the preceding claims, The schedule (10) includes executable control code for the light control system, and the changes in step e. (110) are made by executing the control code.

6. The method according to any one of the preceding claims, In step d. (107, 108), the schedule (10) is created by means of inference through the trained machine learning model (5) when the at least one subtask (4) is input.

7. The method according to any one of the preceding claims, In step b. (102, 103), at least one subtask (4) is created taking into account the context.

8. The method according to any one of the preceding claims, In step b. (102, 103), the at least one subtask (4) is created; in step c. (104, 105, 106), the at least one lighting object (7) and at least one lighting feature (6) of the determined at least one lighting object (7) are determined by means of inference through the trained machine learning model (5); and / or in step d. (107, 108), the schedule (10) is created by means of inference through the trained machine learning model (5) taking into account at least one additional factor.

9. The method according to claim 8, The additional factors mentioned above are musical works, location features, the additional state of at least one additional device, light shows, and / or sketches.

10. The method according to any one of the preceding claims, In step c. (104, 105, 106), at least one signal from at least one sensor and / or at least one transmitter is acquired, and in step c. (104, 105, 106), the at least one lighting object (7) is determined by inference through the trained machine learning model (5) upon input of the at least one signal, and / or in step d. (107, 108), the schedule (10) is created by inference through the trained machine learning model (5) upon input of the at least one signal.

11. The method according to any one of the preceding claims, The voice command (2) is a control code of at least one first naming scheme, and step b. (102, 103) and / or step c. (104, 105, 106) includes converting the control code from the at least one first naming scheme to a second naming scheme.

12. The method according to any one of the preceding claims, The method described herein includes, in step b. (102, 103): Evaluate the voice command (2) and / or the at least one subtask (4), and if the verification has a negative result, request the input of another voice command (2).

13. The method according to any one of the preceding claims, Among them, it is feasible to obtain at least one setting of the at least one lighting state by calling the database (8).

14. A computer system (20) configured to implement one of the methods according to claims 1 to 13, the computer system comprising at least one interface (21) for obtaining the voice instruction (2) to implement step a. (101), a computer-readable storage medium (22) for storing the voice instruction (3) and the trained machine learning model (5) to implement steps b. to d. (102, 103, 104, 105, 106, 107, 108), and at least one data processing means (23) configured to implement steps b. (102, 103) and d. (107, 108).

15. A computer program product comprising software code segments configured to enable implementation of the method according to any one of claims 1 to 13 by at least one processor.

Images