Lighting control apparatus, corresponding method and computer program product
View Patent ↗A lighting control apparatus for producing control signals or cues for controlling operating parameter of one or more controlled lighting devices, including a learning machine configured for performing: a supervised learning phase, wherein, as a function of a first set of audio and/or video files coupled with a first set of control signals, the machine produces mapping rules between the audio and/or video files and the control signals of these first sets; an unsupervised learning phase wherein the machine receives a second set of audio and/or video files and produces, from the second set of audio and/or video files, a second set of control signals as a function of the mapping rules. The learning machine may be configured to carry out a reinforced learning phase with the production of an evaluation ranking of the mapping rules and the possible elimination of mapping rules.
1. A lighting control apparatus for producing control signals of at least one operating parameter of at least one controlled lighting device, wherein the apparatus includes a learning machine configured for performing:
a supervised learning phase, wherein, as a function of a first set of audio and/or video files coupled with a first set of control signals, the machine produces mapping rules between the first set of audio and/or video files and the first set of control signals;
an unsupervised learning phase wherein the machine receives a second set of audio and/or video files and produces, from the second set of audio and/or video files, a second set of control signals as a function of the mapping rules.
2. The lighting control apparatus of claim 1 , wherein the learning machine is configured for performing a reinforced learning phase by producing an evaluation ranking of the mapping rules.
3. The lighting control apparatus of claim 2 , wherein the learning machine is responsive to usage feedback signals and is configured for producing the evaluation ranking of the mapping rules as a function of the usage feedback signals.
4. The lighting control apparatus of claim 3 , wherein the learning machine is responsive to usage feedback signals selected out of:
usage evaluation signals provided to the learning machine, and
usage frequency signals detected by the learning machine and indicating a frequency of selection of certain audio and/or video files coupled with respective control signals.
5. The lighting control apparatus of claim 2 , wherein the learning machine is configured for deleting mapping rules in tail positions in the evaluation ranking.
6. The lighting control apparatus of claim 1 , wherein the learning machine is configured for producing the first set of audio and/or video files coupled with a first set of control signals via a clustering procedure applied to combinations of audio and/or video file with control signals coupleable with the audio and/or video files.
7. The lighting control apparatus of claim 1 , wherein the learning machine includes a finite state machine, having a plurality of machine states wherein:
a transmission function f produces a transition from a state S(t) in the plurality of machine states to a subsequent state S(t+1) in the plurality of machine states as a function of an input value I(t), and
a transmission function g couples a state S(t) in the plurality of machine states to an output value O(t) as a function of an input value I(t),
wherein the input values I(t) and output values O(t) are indicative of audio and/or video files and control signals, respectively, linked via the mapping rules.
8. The lighting control apparatus of claim 7 , wherein the transmission function f of the finite state machine produces a transition from a state S(t) in the plurality of machine states to a subsequent state S(t+1) as a function of a probability P that an event associated with the subsequent state S(t+1) in the plurality of machine states occurs after the state S(t) in the plurality of machine states.
9. The lighting control apparatus of claim 7 , wherein the transmission function f of the finite state machine is configured for correlating, in the unsupervised learning phase, the output values O(t) with a sequence of N previous output values and with a sequence of M previous input values.
10. The lighting control apparatus of claim 1 , wherein the learning machine includes:
an artificial neural network configured for learning, during the supervised learning phase, the mapping rules between the first set of audio and/or video files and the first set of command signals, and
a long short term memory configured for learning, during the unsupervised learning phase, the evolution over time of the control signals in the second set of control signals produced as a function of the mapping rules.
11. The lighting control apparatus of claim 10 ,
wherein the learning machine is configured for performing a reinforced learning phase by producing an evaluation ranking of the mapping rules,
wherein the long short term memory is configured for learning, during the reinforced learning phase, the evolution over time of the control signals in the second set of control signals produced as a function of the mapping rules for which the evaluation ranking has been produced.
12. A method of controlling at least one operating parameter of at least one lighting device, the method including:
providing a lighting control apparatus, wherein the lighting control apparatus includes a learning machine configured for performing:
a supervised learning phase, wherein, as a function of a first set of audio and/or video files coupled with a first set of control signals, the machine produces mapping rules between the first set of audio and/or video files and the first set of control signals;
an unsupervised learning phase wherein the machine receives a second set of audio and/or video files and produces, from the second set of audio and/or video files, a second set of control signals as a function of the mapping rules;
supplying the lighting control apparatus with a first set of audio and/or video files coupled with a first set of control signals,
activating the learning machine in the system, and
controlling the at least one lighting device with the control signals produced by the learning machine.
13. A non-transitory computer readable medium, loadable in a memory of a learning machine in a lighting control apparatus, wherein the learning machine is configured for performing:
a supervised learning phase, wherein, as a function of a first set of audio and/or video files coupled with a first set of control signals, the machine produces mapping rules between the first set of audio and/or video files and the first set of control signals;
an unsupervised learning phase wherein the machine receives a second set of audio and/or video files and produces, from the second set of audio and/or video files, a second set of control signals as a function of the mapping rules;
the non-transitory computer readable medium including software code portions for activating the learning machine in a method of controlling at least one operating parameter of the at least one lighting device, the method including:
providing the lighting control apparatus;
supplying the lighting control apparatus with a first set of audio and/or video files coupled with a first set of control signals,
activating the learning machine in the system, and
controlling the at least one lighting device with the control signals produced by the learning machine.
14. The lighting control apparatus of claim 8 , wherein the transmission function f of the finite state machine is configured for correlating, in the unsupervised learning phase, the output values O(t) with a sequence of N previous output values and with a sequence of M previous input values.