Ai assisted dynamic modification of musical structure for attaining target mental state via audio-based operant conditioning

Abstract

A method for providing real-time music stimuli to urge a subject having a time-varying mental state to attain a target state includes receiving a first physiological-signal that carries information that represents the time-varying mental state, using the physiological signal and information about the target state to generate a music stimulus in real time and transmitting the music stimulus for exposure thereof to the subject.

Description

Attorney Docket No. 30263-008W01Al ASSISTED DYNAMIC MODIFICATION OF MUSICAL STRUCTURE FOR ATTAINING TARGET MENTAL STATE VIA AUDIO-BASED OPERANT CONDITIONING FOR BRAIN TRAININGCross-References to Related Applications

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 703,255 filed October 4, 2024, the entire contents of which are incorporated herein by reference.Background of the Invention

[0002] This invention relates to operant conditioning.

[0003] An electroencephalogram is typically dominated by frequencies within a particular band of frequencies. Both the dominance of a particular range of frequencies and the relationships between bands of frequencies have been found empirically to be associated with different states of consciousness. For example, an electroencephalogram dominated by very low frequencies, on the order 0.1 Hz to 3 Hz, is typically associated with sleep or a sense of detached awareness. Such an electroencephalogram is said to be dominated by “delta waves.” One dominated by frequencies between about 5 Hz and 15 Hz has been found to be associated with a sense of relaxation, creativity, and visualization. This range if frequencies is often referred to as the “alpha waves.”

[0004] It is generally desirable for a human to be able to switch between these various states of consciousness on demand. For instance, the ability to switch into a state dominated by delta waves on cue would be desirable for quickly falling asleep.

[0005] Unfortunately, many humans find it difficult to switch between brain states on demand. Those in the creative professions may find themselves waiting to reach a creative state and attempting all manner of peculiar ritual in an effort to arrive there sooner.

[0006] In some cases, this lack of control over one’s state of consciousness is sufficiently extreme such that pharmaceuticals are used to artificially induce the desired state of consciousness.Summary

[0007] In one aspect, the invention features a method including causing a local circuit and a remote circuit that are interconnected by a network to cooperate in using real-Attorney Docket No. 30263-008W01 time music stimuli to urge a subject having a time-varying state to attain a target state. The remote circuit including an adaptation engine and a personalization engine. The target state and the time- varying state are mental states of the subject. The method includes doing so by having the remote circuit’s personalization engine receive, from the local circuit, an electrical signal that carries information indicating the target state and having it receive a first physiological-signal from the local circuit. This first physiological-signal is an electrical signal that carries information that represents the time-varying mental state at a first time. The method continues with causing the adaptation engine to generate a first music stimulus based at least in part on the first physiological-signal and to transmit the first music stimulus to the local circuit for exposure thereof to the subject. Then, the personalization engine receives a second physiological-signal from the local circuit. This second physiological-signal is an electrical signal that carries information that represents the time- varying mental state at a second time, which follows the first time. The personalization engine then determines, based at least in part on the second physiological-signal that the subject’s mental state continues to differ from the target state. The method then continues with causing the adaptation engine to generate a second music stimulus based at least in part on the second physiological-signal and to transmit the second music stimulus to the local circuit for exposure thereof to the subject. The personalization engine then receives a third physiological-signal from the local circuit. This third physiological- signal is an electrical signal that carries information that represents the time-varying mental state at a third time, which follows the second time. The personalization engine then determines, based at least in part on the third physiological-signal, that the subject’s mental state has moved toward from the target state.

[0008] Practices of the method include those in which generating the first music stimulus includes transforming musical structure of selected music based at least in part on features present in the first physiological-signal and those in which generating the first music stimulus includes using a generative model to generate the first music stimulus based at least in part on features present in the first physiological-signal.

[0009] In some practices of the method, the first physiological-signal includes an electroencephalogram having a first spectrum and a second spectrum. The first spectrum consists of frequencies within a first band of frequencies and the second spectrum consists of frequencies within a second band of frequencies, with these bands being disjoint so that they have no frequencies in common. Stated differently, given four real-valued frequencies fl, f2, f3, and f4, there exists a first band defined by [fl, f2] and a second band defined by [f3, f4] where f3 > f2. Among these practices are those in which generating the first music stimulus is carried out based at least inAttorney Docket No. 30263-008W01 part on a change in a ratio of power present in the first spectrum and power present in the second spectrum and those in which generating the first music stimulus is carried out based at least in part on a change in power present in the first spectrum.

[0010] In other practices, receiving the first and second physiological-signals includes receiving the signals at the remote circuit after the signals have been transmitted to the remote circuit by the local circuit. In these practices, the local circuit being one of a plurality of local circuits, each of which receives physiological signals from a different subject and each of which provides the physiological signals to the remote circuit.[Oil] Still other practices include those in which generating the first music stimulus includes generating the first music stimulus based at least in part on an event signal that indicates occurrence of an external event that affects the subject’s time-varying state. In such practices, the first stimulus is generated based at least in part on both the event signal and on the first physiological-signal.

[0012] Also among the practices of the invention are those in which generating the first music stimulus includes selecting two of the spectra that are constituents of the electroencephalogram and determining the time-varying state at a particular based on a power ratio between the two of the spectra. These spectra consist of frequencies within a corresponding band of frequencies that are not in any other band of frequencies.

[0013] Some practices include, for each spectrum among the spectra of the first physiological-signal, identifying the frequency that consumes the largest share of power within that spectrum. Among these are practices in which the first physiological-signal includes an electroencephalogram that includes the spectra.

[0014] Practices further include those in which generating the first music stimulus includes, for each spectrum among the spectra that constitute the first physiological signal, using a spectral-entropy analyzer to output a level of entropy in the spectrum. Among these practices are those in which the first physiological-signal includes an electroencephalogram.

[0015] Still other practices include those in which determining, based at least in part on the second physiological-signal that the subject’s mental state continues to differ from the target state includes providing information from the second physiological signal to a state-detection layer that incorporates a classifier for classifying the subject’s mental state. Among these practices are those in which the classifier is a neural network and those in which it is a support vector machine.Attorney Docket No. 30263-008W01

[0016] In still other practices, determining, based at least in part on the second physiological-signal, that the subject’s mental state continues to differ from the target state includes providing information from the second physiological signal to a statedetection layer and causing the state-detection layer to identify a shift in emotional valence of the time-varying state of the subject.

[0017] Other practices include those in which causing the adaptation engine to generate the second music stimulus based at least in part on the second physiological- signal includes transforming musical structure of an existing musical composition. Among these are practices are those in which the adaptation engine to transform an existing musical composition’s rhythmic structure, those in which the adaptation engine transforms harmonic structure of an existing musical composition, and those in which the adaptation engine transforms spectral content of the existing musical composition.

[0018] In still other practices, causing the adaptation engine to generate the second music stimulus based at least in part on the second physiological-signal includes causing the adaptation engine to add a reward stimulus to an existing music composition and to use the existing music composition, with the reward stimulus having been added thereto, as the second music stimulus.

[0019] Practices further include those in which causing the adaptation engine to generate a first music stimulus includes transforming a rhythmic density of an existing musical composition or transforming the spectral brightness of an existing musical composition.

[0020] In another aspect, the invention features an apparatus for use by a plurality of subjects, each of the subjects having a time-varying state and being equipped with a corresponding local circuit from a plurality of local circuits, each of the time-varying states being a mental state of a corresponding one of the subjects from the plurality of subjects. Such an apparatus includes a remote circuit that is in communication with each of the local circuits from the plurality of local circuits over a network. The remote circuit includes an adaptation engine and a personalization engine. The personalization engine is configured: to receive physiological signals from each of the subjects, to receive target states from each of the subjects, each of the target states having been specified by a corresponding one of the subjects, and to provide, to the adaptation engine, information indicative of a current state of each of the subjects. The adaptation engine is configured: to provide music stimuli for each of the subjects based at least on part on one or more physiological signals from each of the subjectsAttorney Docket No. 30263-008W01 and to provide the music stimuli to corresponding ones of the local circuit for exposure to corresponding ones of the subjects.

[0021] In another aspect, the invention features distributed feedback-circuitry that uses operant conditioning to train a subject to achieve a target state of consciousness. The distributed feedback-circuitry includes biofeedback circuitry that includes a feature-extraction circuit and a controller. The feature-extraction circuit receives a real-time measurement signal that carries information concerning one or more physiological signals and generates a measured feature-set and a target feature-set therefrom. The target feature-set includes features for achieving the target state of consciousness. The controller causes transmission of a conditioning stimulus to be listened to by the subject. This conditioning stimulus causes the measured feature-set to be driven towards the target feature-set. The controller causes the conditioning stimulus to include a conditioning- audio stimulus that transitions between being a base-audio stimulus that lacks a reward stimulus and a base-audio stimulus that has been operated on to incorporate a reward stimulus therein. The controller causes this transition based on progress made in causing the measured feature-set to conform to the target feature-set and also causes an artificial-intelligence engine to generate the base-audio stimulus using a music model that has been trained on a music dataset to generate music that has features that cause the measured feature-set to be driven towards the target feature-set.

[0022] In some embodiments, the subject is an individual, whereas in others, the subject is a group. Accordingly, in the latter case, the one or more physiological signals is averaged over the group. Examples of physiological signals include those representative of heart rate and / or heart-rate variability. Other examples include signals indicative of brain or nervous system activity, including electroencephalograms. In such embodiments, the artificial-intelligence engine generates the base-audio stimulus to include particular frequencies or tracks that are based on trends in the physiological signals, including trends in the heart rate, heartrate variability, and electroencephalograms.

[0023] Among the embodiments in which the artificial-intelligence engine is configured to determine trends in the physiological signals of a group as a whole are those in which the artificial-intelligence engine uses those group trends in conjunction with pre-determined algorithms to cause the base-audio signal to include one or more of appropriate tracks, appropriate frequencies and combinations thereof, and binaural beats, thereby urging the subjects, including individual members of the group, to be driven into the target state of consciousness and, in some embodiments, to cause those individual members to be urged towards a target physiological state or, if theAttorney Docket No. 30263-008W01 individual members are already in such a state, to deepen or extend that state. Such pre-determined algorithms are determined empirically, essentially by exposing test subjects to particular stimuli, recording the effect of those stimuli, and inferring a cause- and-effect relationship between the exposure and the effect of that exposure, including the use of statistical tests to determine the confidence level of the hypothesis that those stimuli were the cause of the relevant observed effect.

[0024] In still other embodiments, the music model is one that has been trained to reward a movement towards the target state by triggering the inclusion of auditory or verbal cues. In still other embodiments, there exists a visual model that has been trained such that movement towards the target state triggers a visual cue as a reward for such movement.

[0025] It is possible to derive certain parameters from the physiological signals, such as the dominant frequencies, amplitudes, and variations in power as a function of time, i.e., power dynamics, within various bands of an electroencephalograph, such as the delta band, the theta band, the alpha band, and the gamma band. Other parameters that can be derived from physiological signals include coherence and phase changes.

[0026] In some embodiments, the artificial-intelligence engine is configured to take such parameters into account to add certain predetermined auditory triggers, such as frequency content, sounds, various instruments, each of which can be viewed as a contributor of a characteristic frequency distribution that defines a particular timbre thereof, melodies, harmonic content, including chord progressions, changes in overall tempo, changes in rhythmic patterns, including such features as synthetic rubato generation, changes in the number of beats per measure, changes in the duration of quarter notes, tempo changes, or rhythm changes. Such parameters include those derived from heart rate, heart-rate variability, respiration rate, respiration volume, galvanic skin response, and movement. As a result, the controller, through the assistance of the artificial-intelligence engine causes a conditioning-audio stimulus in whatever genre or style is required based on real-time live physiological dynamics of the subject.

[0027] A system along the lines of the foregoing is usable in a variety of environments, such as in the subject’s home or automobile. For example, a suitable equipped smart car or smart house would include sensors that provide the system with information indicative of increased agitation or fatigue. In such cases, the system responds by causing more uplifting music to be played within the environment. In some embodiments, the system changes environmental features in addition to the audio environment. Examples includes those in which the artificial-intelligenceAttorney Docket No. 30263-008W01 engine relies on a model that has been trained to cause changes in lighting, such as changes in luminosity or color. Other examples include those in which the artificialintelligence engine has been relies on a model that has been trained to control ambient animations or videos in the environment in an attempt to use operant conditioning to urge the subject to arrive at a target state of consciousness. Among the embodiments are those in which music model includes a neural network, those in which the music model is one that has been trained to compose music having characteristics that cause the measured feature-set to be driven towards the target feature-set, those in which the music model is one that has been trained to include binaural beats in music, the binaural beats having been selected to cause the measured feature-set to be driven towards the target feature-set, those in which the music model is one that has been trained to compose music that includes frequency modulation, the frequency modulation having been selected to cause the measured feature-set to be driven towards the target feature-set, those in which the music model is one that has been trained to compose music that includes an embedded frequency, the embedded frequency having been selected to cause the measured feature-set to be driven towards the target feature-set, those in which the music model is one that has been trained to add a baseline to music, the bassline having been selected to cause the measured feature-set to be driven towards the target feature-set, those in which the music model is one that has been trained to add a superposition of pulse trains to music, the pulse trains having been selected to cause the measured feature-set to be driven towards the target feature-set, those in which the music model is one that has been trained to add a stem to music, the stem having been selected to cause the measured feature-set to be driven towards the target feature-set, those in which the music model is one that has been trained to add a stem to music and to synchronize the stem with the music, the stem having been selected to cause the measured feature-set to be driven towards the target feature-set, and those in which the music model is one that has been trained to filter music through a passband, the passband having been selected to cause the measured feature-set to be driven towards the target feature-set.

[0028] Also among the embodiments are those in which the music model is one that has been trained to suppress a set of frequencies in the music, the set of frequencies including all frequencies above a first frequency, the first frequency having been selected to cause the measured feature-set to be driven towards the target feature-set, those in which the music model is one that has been trained to suppress a set of frequencies in the music, the set of frequencies including all frequencies below a first frequency, the first frequency having been selected to cause the measured feature-set to be driven towards the target feature-set, and those in which the music model is one that has been trained to suppress a set of frequencies in the music, the set ofAttorney Docket No. 30263-008W01 frequencies being a union of a first set, which includes all frequencies below a first frequency, and a second set, which includes all frequencies above a second frequency, the second frequency being greater than the first frequency, the first and second frequencies having been selected to cause the measured feature-set to be driven towards the target feature-set.

[0029] In some embodiments, the feature-extraction circuit receives information from a real-time electroencephalogram.

[0030] In other embodiments, the feature-extraction circuit receives information representative of one or more of real-time heart rate, heart-rate variability, respiration rate, body temperature, peripheral body temperature, galvanic skin response, changes in galvanic skin response, changes in muscle tension, respiration rate, and respiration volume.

[0031] Some embodiments include a robot, with the biofeedback circuitry being in whole or in part incorporated within the robot.

[0032] Still other embodiments include those in which the distributed feedbackcircuitry includes local circuitry that provides the biofeedback circuitry with information concerning real time physiological signals and wherein the biofeedback circuitry is accessible to the local circuitry via a network.

[0033] In another aspect, the invention features a method that includes causing a measured feature-set to be driven towards a target feature-set by receiving, from a subject, real-time physiological information and information that is indicative of a target state-of-consciousness, based on features in the real-time physiological information, forming a conditioning stimulus to cause a state-of-consciousness of the subject, as determined based on the real-time physiological information, to move towards the target state-of-consciousness, and transmitting the conditioning stimulus to the subject. In such a method, forming the conditioning stimulus includes using an artificial-intelligence engine including a music model to generate music having features that are selected to drive the measured feature- set towards the target featureset. The method is limited to only those implementations that are non-abstract.

[0034] In some embodiments, the biofeedback circuitry includes application-specific circuitry that includes resistors, capacitors, inductors, transistors, and diodes together with a clock that controls the intervals in which charge is made to move through the various circuit elements. Among the circuit elements are arrays of semiconductor devices that maintain one of two desired states over time and that are made to transition between states at selected times.Attorney Docket No. 30263-008W01

[0035] The various steps carried out by the biofeedback circuitry have proven to be incapable of being performed in a human mind given its current state of evolution. Indeed, it was for this reason that biofeedback circuitry was required to implement the methods described herein.

[0036] Additionally, the various steps carried out by the biofeedback circuitry have proven to be incapable of being performed have also been found to be incapable of being on a generic computer. Thus far, they have only been performed on a nongeneric computer. Based on Supreme Court precedent, there exist two and only two types of digital computers: those that are “generic” and those that are not. As used herein, a ri'non-generic” computer means a computer that is not “generic” as that term has been defined by the courts of the United States as of the filing date of this application.

[0037] All attempts to cause the biofeedback circuitry to perform the methods described herein in an abstract manner have thus far failed. Each attempt resulted in performance of the method in a non-abstract manner, where “non-abstract” is defined herein as the converse of “abstract” as that term is used by the Supreme Court of the United States.

[0038] The claims are explicitly defined to include only non-abstract implementations of the recited apparatus and methods, where “non-abstract” has been defined as above. Any party who presumes to construe the claims as being abstract in nature would simply be proving that it is possible to improperly construe the claims in a manner inconsistent with express statements to the contrary within the specification.

[0039] One may be inclined to believe that the various steps carried out by the remote circuitry are simply mental steps that can be carried out in the human mind. However, this is merely one of a long string of beliefs that have been debunked by the scientific method. Based on controlled experiment, it has been found that the human mind in its current state of evolution is unable to perform the steps carried out by the remote circuitry. Indeed, it was for this reason that remote circuitry was required to implement the methods described herein. After all, if the human mind could in fact carry out the steps, there would have been no need to bother with constructing the relevant remote circuitry as described herein.

[0040] Additionally, the various steps carried out by the remote circuitry have proven to be incapable of being performed have also been found to be incapable of being on a generic computer. Thus far, they have only been performed on a non-generic computer. Based on Supreme Court precedent, there exist two and only two types of digital computers: those that are “generic” and those that are not. As used herein, aAttorney Docket No. 30263-008W01“non-generic” computer means a computer that is not “generic” as that term has been defined by the courts of the United States as of the filing date of this application.

[0041] All attempts to cause the remote circuitry to perform the methods described herein in an abstract manner have thus far failed. Each attempt resulted in performance of the method in a non-abstract manner, where “non-abstract” is defined herein as the converse of “abstract” as that term is used by the Supreme Court of the United States.

[0042] The claims are explicitly defined to include only non-abstract implementations of the recited apparatus and methods, where “non-abstract” has been defined as above. Any party who presumes to construe the claims as being abstract in nature shall be regarded as simply having demonstrated that it is not impossible to improperly construe the claims in a manner inconsistent with express statements to the contrary within the specification.

[0043] Other features and advantages of the invention are apparent from the following description, and from the claims.Description of Drawings

[0044] FIG. 1 shows a brain-training circuit providing a subject with dynamically- adapted music in real time in response to a physiological signal from the subject;

[0045] FIG. 2 shows details of the adaptation engine of FIG. 1 ;

[0046] FIG. 3 shows details of the personalization engine of FIG. 1 ;

[0047] FIG. 4 shows steps in a method carried out by the circuit shown in FIG. 1; and

[0048] FIG. 5 shows an alternative embodiment of the adaptation engine of FIG. 1.

[0049] FIG. 1 shows distributed feedback-circuitry comprising a headset;

[0050] FIG. 2 is a method carried out by the distributed feedback-circuitry of FIG. 1 ; and

[0051] FIG. 3 is an alternative embodiment of the distributed feedback-circuitry of FIG. 1.Description

[0052] FIG. 1 shows a distributed brain-training circuit 10 having a local circuit 12 and a remote circuit 14 that cooperate to use operant conditioning to urge a subject 16 from a subject set 18 to transition into a target state and to remain in that target stateAttorney Docket No. 30263-008W01 upon attainment thereof. As used herein, a subject’s “state” is that subject’s mental state; a “target” state is a “state” that the subject 16 designates as being a state that the user would like to attain through the use of the brain-training circuit.

[0053] A “target state” is one of: a target emotional state, a target cognitive state, and / or a target psychological state. Examples of a target state include: a relaxed state, a stress-reduced state, a state conducive to sleep, a state conducive to enhancement of focus, and a state conducive to greater attention. Other examples of a target state include a state of mindfulness, an energetic state, a happy state, and a state of flow. Still other examples of a target state are a state of having a positive derivative of relaxation with respect to time, a state of having a positive derivative of focus with respect to time, and a state of having a positive derivative of cognitive flexibility with respect to time.

[0054] The brain-training circuit 10 relies on receiving a physiological signal 20 from the subject 16. This physiological signal 20 is indicative of the user’s current state. In the illustrated embodiment, a wearable platform 22 carries a transducer set 24 having one or more transducers 26. The transducer set 24 obtains various measurements and provides the physiological signal 20 to the brain-training circuit 10. This physiological signal 20 incorporates information from those measurements. Examples of a “signal” include a scalar- valued signal and a vector- valued signal in which each element is a scalar value.

[0055] In response to the physiological signal 20, the brain-training circuit 10 provides a music stimulus 28 to the subject 16. The brain-training circuit 10 dynamically adapts this music stimulus 28 in real time in an effort to guide the subject 16 towards the target state. As used herein, “real time” means within a time frame that is short enough to be perceived by one of ordinary skill in the art as immediate or timely for its intended purpose. In the illustrated embodiment, the particular stimulus comprises music that comprises musical structure and the brain-training circuit 10 dynamically adapts the music stimulus 28 by transforming the musical structure.

[0056] FIG. 1 shows a particular embodiment in which the transducer set 24 comprises an electroencephalograph 30 having plural electrodes that are contact with the subject’s scalp at one or more locations via one or more electrodes. In this embodiment, the physiological signal 20 is an electroencephalogram. As such, the physiological signal 20 directly indicates the subject’s brain activity.

[0057] Other examples of transducer sets 24 include those that indirectly indicate the subject’s brain activity. Embodiments include those in which the transducer set 24 includes one or more of: the electroencephalogram 30, a galvanometer 32 thatAttorney Docket No. 30263-008W01 measures galvanic skin response and changes thereof; an electromyograph 34 that monitors changes in muscle tension; a spirometer 35 that measures respiration rate or respiration volume; a heart-rate monitor 36, which provides information on heart rate, changes in heart rate, and heart-rate variability; and a temperature monitor 38 for measuring body temperature, including peripheral body temperature.

[0058] The physiological signal 20 is then transmitted in digitally-encoded form to the local circuit 12, which executes a trainer 40. The local circuit 12 includes an integrated circuit 42 that has been specially configured to execute the trainer 40, a timing circuit 44 that provides a periodic signal to the integrated circuit 42, and a power supply 46 that convers a variable input voltage into a stable voltage for use by the integrated circuit 40.

[0059] Examples of a local circuit 12 include circuitry embedded in various portable and / or wearable devices, such as a smartphone, a tablet, smart jewelry, a smart watch, and a laptop. Other examples of a local circuit 12 include non-portable devices, such as a personal computer. A suitable implementation of trainer 40 is an application that executes on the local circuit 12.

[0060] In operation, the subject 16 communicates a target state to the trainer 40. The brain-training circuit 1 attempts to urge the subject’s state towards this target state based at least in part on information from the physiological signal 20. It does so by providing the subject 16 with a music stimulus 28 that has been dynamically adapted in real time to the subject’s state.

[0061] The trainer 40 displays information from the physiological signal 20 in real time on the local circuit 12 for the subject’s inspection. In addition, the trainer 40 causes the local circuit 12 to forward a feature signal 48 to the remote circuit 14. The feature signal 48 comprises information derived from the physiological signal.

[0062] In those cases in which the signal transducer set 24 comprises an electroencephalograph 30, the feature signal 48 carries such information as information concerning changes in amplitude or power of any frequency or combination of frequencies present in an electroencephalogram provided by the electroencephalograph 30. Other examples of such information include information concerning changes in amplitude, or power, in one or more of alpha, beta, gamma, delta, and theta waves from the foregoing electroencephalogram. Other examples include changes in the ratio of power present in any two or more of alpha, beta, gamma, delta, and theta waves from the foregoing electroencephalogram.Attorney Docket No. 30263-008W01

[0063] Still other examples of information carried by the feature signal 48 include information concerning changes in coherence or phase between various components of the foregoing electroencephalogram and information concerning changes in the spectrum of frequency -domain representation of the electroencephalogram, such as changes in ratios of power in pairs of bands or frequencies in the foregoing electroencephalogram. In those cases in which the transducer set 24 comprises electrodes that are in contact with different parts of the subject’s scalp, another example of such information include changes in in left / right symmetry or in the symmetry of the electroencephalograms obtains at those different locations.

[0064] In some embodiments, the remote circuit 14 executes on cloud-based hardware 50. In such embodiments, the remote circuit 14 is shared by plural instances of the local circuit 12 via a network. As a result, the remote circuit 14 provides services to other subjects 16 from the subject set 18, all of whom are all attempting to attain possibly different target states. As a result, the remote circuit 14 is available for concurrent use by additional subjects 16 (from the subject set 18) who interact with the remote circuit 14 in the same manner using corresponding wearable platforms 22 and local circuits 12.

[0065] In other embodiments, the remote circuit 14, although separate from the local circuit 12, is used only by that local circuit 12. An example of such an embodiment is one in which the remote circuit 14 is part of a robot. An example of such a robot is one manufactured by HANSEN ROBOTICS that has been given the tradename “SOPHIA.” In still other embodiments, the hardware that executes the remote circuit 14 is distributed between a robot and cloud-based hardware 50.

[0066] The remote circuit 14 includes a personalization engine 52 and an adaptation engine 54, the former being configured to receive target-state information 56 that the local circuit 12 transmitted in response to an instruction from the trainer 40. This target-state information 56 comprises information indicative of what sort of target state the subject 16 wishes to attain.

[0067] The personalization engine 52 extracts features from the feature signal 48. This results in a measured feature-set 58 that includes information concerning the subject’s real-time state. Based on the target-state information 56, the personalization engine 52 defines a target feature-set 60. It then provides both the target feature-set 60 and the measured feature- set 58 to the adaptation engine 54.

[0068] Referring now to FIG. 2, the adaptation engine 54 includes a controller 62 that uses the target feature-set 60 and the measured feature-set 58 as a basis for providing a control signal 54 to a conditioner 66.Attorney Docket No. 30263-008W01

[0069] The conditioner 66 comprises an audio generator 68, a video generator 70, and a reward generator 72. The audio generator 68 provides a base-audio stimulus 74; the video generator 70 provides a video stimulus 76; and the reward generator 72 provides a reward stimulus 78. The base-audio stimulus 74 and the occasional reward stimulus 78 combine to form a conditioning-audio stimulus 80. The conditioningaudio stimulus 80 is then combined with the video stimulus 76 to form a combined stimulus 82. It is this combined stimulus 82 that is ultimately provided to the subject 16. This combined stimulus 82 provides positive or negative reinforcement to the subject 16 in an attempt to condition the subject 16 to reach the target feature-set 60.

[0070] The existence of positive or negative reinforcement in the combined stimulus 82 arises as a result of the adaptation engine 54 having provided a reward signal 84 to control a reward switch 86. Depending on the state of the reward switch 86, a first adder 88 either combines the reward stimulus 78 with the base-audio stimulus 74 and removes the reward stimulus 78 from the base-audio stimulus 74. This controls whether the resulting conditioning-audio stimulus 80 provides positive reinforcement or negative reinforcement to the subject 16. A second adder 90 receives the conditioning-audio stimulus 80 and combines it with the video stimulus 76 to form the combined stimulus 82.

[0071] The reward signal 84 provided by the adaptation engine 54 thereby controls whether or not the combined stimulus 82 will comprise a superposition of the baseaudio stimulus 74 and the reward stimulus 78 or just the base-audio stimulus 74 by itself. By controlling whether the reward stimulus 78 is present or absent based on the difference between the target feature-set 60 and the measured feature-set 58, the adaptation engine 54 carries out the conditioning in an effort to drive the measured- feature set 48 closer to the target feature-set 60.

[0072] The reward stimulus 78 changes in response to the progress being made towards arriving at the target feature-set 60. As used herein, a change to the reward stimulus 78 includes causing the reward stimulus 78 to appear, causing the reward stimulus 78 to disappear, and altering the reward stimulus 78. In some embodiments, the reward stimulus 78 is a single tone; in other embodiments, it is a weighted combination of frequencies having complex-valued weighting coefficients; and in still other embodiments, it is a superposition of tones having different wavelengths weighted by different values of magnitude and / or phase.

[0073] The subject’s neurons are thus exposed to the audio portion of the combined stimulus 82 via the auditory pathway. This enables the conditioning process to be carried out. The adaptation engine 54 tailors the combined stimulus 82 to cause theAttorney Docket No. 30263-008W01 measured feature-set 58 to transition into the target feature-set 60. It does so by reducing a difference between the measured feature-set 58, which is arriving at the adaptation engine 54 in essentially real time, and the target feature-set 60, which was provided by the trainer 40 at the beginning of the conditioning process. This difference is reduced at least in part by the reward signal 84, which selectively adds or subtracts the reward stimulus 78 from the combined stimulus 82 as well as by the selection of content the reward stimulus 78 and the content of both the base-audio stimulus 74 and the video stimulus 76 so as to condition the subject’s relevant neurons so that they achieve and maintain a target brain state.

[0074] The adaptation engine 54 transmits the combined stimulus 82 back to the trainer 40 at the local circuit 12. Upon receipt of the combined stimulus 82, the trainer 40 separates the video stimulus 76 from the conditioning-audio stimulus 80 and displays the video stimulus 76 on the local circuit 12. The trainer 40 then sends the conditioning-audio stimulus 80 to the wearable platform 22 to be listened to by the subject 16 through loudspeakers that are a constituent part of the wearable platform 22. As a result, the subject’s neurons are exposed to the combined stimulus 82 using two different sensory pathways.

[0075] Like the conditioning-audio stimulus 80, the video stimulus 76 reflects the real-time electroencephalogram measurements obtained from the subject 16. Accordingly, the details of the video stimulus 76 change in response to changes in those electroencephalogram measurements. Preferably, the video stimulus 76 and the conditioning-audio stimulus 80 are synchronized. As a result, when the conditioningaudio stimulus 80 is changed, the video stimulus 76 sustains a corresponding change.

[0076] The trainer 40 continues to receive measurement signals from the transducer set 24 as the subject 16 is exposed to the combined stimulus 82. These are transmitted to the personalization engine 52 to serve as a basis for feedback control over the process of guiding the subject 16 towards the target state.

[0077] The personalization engine 52 carries out further feature extraction on the feature signal 48. The resulting updated measured feature-sets 48 provide a basis for evaluating the effect of the combined stimulus 82 and, in particular, the progress made towards driving the measured feature-set 58 towards the target feature-set 60. In response to the assessment of such progress by the personalization engine 52, the adaptation engine 54 causes the conditioner 66 to formulate a revised combined stimulus 82. It then transmits the revised combined stimulus 82 back to the trainer 40 so that the neurons that are to be conditioned can be exposed to them via the subject’s sensory pathways.Attorney Docket No. 30263-008W01

[0078] The adaptation engine 54 modifies one or more compositional aspects of music in response to detected changes in the feature signal 48. As used herein, “compositional aspects” includes the music’s musical structure. Among the transformations include the incorporation of specific precomposed tracks that incorporate certain spectral features therein. These spectral features have been selected to function as a reward stimulus 78 for a detected change in the physiological signal 20. In some cases, the precomposed track is one that has been precomposed by artificial intelligence system 84.

[0079] As used herein, causing music to dynamically adapt means transforming the music’s musical structure in real time. Such transformations exclude changes that do not affect musical structure, such as changing the music’ s playback volume, stopping the music, and starting the music. Examples of aspects of the musical structure include rhythmic structure, harmonic structure, and spectral content.

[0080] In effect, the combination of the adaptation engine 54 and the personalization engine 52 implements a virtual conductor who infers the subject’s state based on the measured feature-set 58 and who, in response to the state inferred therefrom, dynamically interprets the music in a manner that urges the subject 16 into the target state through application of principles of operant conditioning. Such a conductor in effect provides a bespoke listening experience that is finely tuned to the real-time state being experienced by the subject 16. Such interpretation includes transforming features of the musical structure, including one or more of the music’s rhythmic structure, its harmonic structure, and its spectral content.

[0081] Examples of transformations carried out by the combination of the adaptation engine 54 and the personalization engine 52 include transforming the music's rhythmic structure, for example, by changing the music’s tempo and / or changing the concentration of rhythmic events per unit time, i.e., the music’s “rhythmic density” and / or its “regularity.”

[0082] Examples of transforming the music’s harmonic structure include transforming the music’s harmonic progression, modulating the music, transposing one or more portions of the music, and transforming the mode that characterizes the music’s harmony into or out of one of Ionian, Dorian, Phrygian, Lydian, Mixolydian, Aeolian, and Locrian modes.

[0083] Examples of transforming the music’s spectral content include frequencybased changes by transforming one or more of: timbre, instrumentation, and the music’s spectral brightness or darkness.Attorney Docket No. 30263-008W01

[0084] Other examples of frequency-based changes are those that include manipulating amplitude and / or phase of particular bands to cause changes in spatialization and or to cause changes in the perceived direction of sound sources.

[0085] Also among the examples of transforming musical structure are those that include changing the music’s dynamic range or amplitude envelope and changes that include adding special effects, such as reverberation, echo, and / or delay.

[0086] Examples of modifying the music’s structure to form the combined stimulus 82 also include adding a reward stimulus 78.

[0087] In one example, the personalization engine 52 receives a measured feature-set 58 that shows increasing power in the subject’s alpha-wave band. In response, the adaptation engine 54 either transforms the music into a more legato state or incorporates a track having a similar property. Examples of a track include a pre- recorded track and a track generated by an artificial-intelligence engine 92 that is a constituent part of the audio generator 68. The resulting transformation to the music’s structure acts as a reward stimulus 78 to promote an increase in the subject’s own alpha-wave band when, for example, the subject’s target state is that of achieving relaxed wakefulness.

[0088] The artificial-intelligence engine 92 is one that has been trained to select features of musical structure that urge the subject 16 to achieve and / or maintain a target state as specified by the target-state information 56. In some cases, these compositions mimic human musical composition.

[0089] A useful artificial-intelligence engine 92 is one that relies on a music model 94. A music model 94 is a generative model that has been trained using a music dataset 96 that comprises music. A suitable implementation of a music model 94 is a neural network.

[0090] As a result of its having been trained, the music model 94 learns musical patterns, structures, and styles. In some embodiments, the music dataset 96 includes examples of melodies incorporating subtle frequency modulations or binaural beats.

[0091] It has been observed that certain harmonic progressions and certain embedded frequencies promote therapeutic or psychoacoustic effects. Accordingly, in some embodiments, the music model 94 is one that has been trained using a music dataset 96 in which the music incorporates such harmonic progressions and / or embedded frequencies. The music model 94 incorporates these as chords and / or accompaniment in the generated music. In some cases, the music model 94 generates harmonically embedded chords that concurrently complement the generated melody whileAttorney Docket No. 30263-008W01 embedding specific frequencies that have been found to enhance cognitive or emotional responses based on discoveries of such responses through examination of brainwaves.

[0092] In some embodiments, the artificial-intelligence engine 92 embeds binaural beats into the melody. In other embodiments, the artificial-intelligence engine 92 generates music that promotes relaxation, focus, and other desired states by outputting the same note at slightly different pitches, i.e., slightly different frequencies, in different output channels.

[0093] By embedding binaural beats or other auxiliary frequency components within the music, the music model 94 generates compositions with additional layers of depth and complexity. These additional layers potentially enhance the subject’s auditory experience and, most importantly, influence the subject’s state.

[0094] In some cases, the music model 94 is one that has been trained to produce melodies that have been engineered to have enhanced emotion impact. The music model 94 does so by incorporating frequency modulation to create both harmonic and disharmonic effects in the generated music.

[0095] Particularly useful features of the music model 94 include its ability to adapt and improve over time, its fluency in the creation of music according to a variety of styles, and its application across various musical genres and industries.

[0096] Some embodiments also rely on having the artificial-intelligence engine 92 generate individual components of a musical piece, such as melody, harmony, bassline, drums, and other instruments, and to do so in a way that preserves each component as a separate audio track. These components can then be mixed and matched to assemble a generated composition.

[0097] In those embodiments in which the music model 94 has been trained to generate basslines, there exist embodiments in which the music model 94 has been trained on a music dataset 96 that incorporates low-frequency oscillations or sub-bass frequencies that have been selected to interact with binaural beats or other incorporated frequency components in an effort to promote a more immersive experience. Other embodiments feature music models 94 that have been trained specifically to generate basslines incorporate rhythmic low-frequency pulses that align with the beat to reinforce the desired state induced by binaural beats that are present in the music.

[0098] In other embodiments, the music model 94 has been trained on drum patterns, including those that comprise a superposition of pulse trains at different frequencies.Attorney Docket No. 30263-008W01Such superpositions represent drum tracks that include embedded pulse trains at specific frequencies. When subtle enough to avoid overpowering the primary rhythmic elements but not so subtle that they fill to contribute to the overall frequency embedding or superposition of pulse trains in a measurable way, these embedded pulse trains tend to have noticeable psychoacoustic consequences. In these cases, the music model 94 is further trained to ensure that the various drum patterns synchronize with any embedded binaural beats that may be present in a different stem or to ensure that these same drum patterns synchronize with frequency modulations that may be present in different stems. As used herein, the term “stem” is a term of art in music production that refers to an individual audio track or a sub-mix of grouped elements in a musical composition.

[0099] To promote diversity of timbre, it is useful to train the artificial-intelligence engine 92 to provide stems that include additional instrumental parts, each with a different timbre, and to use similar frequency embedding to enhance the auditory effects of the composition as a whole as a result of having incorporated such stems. In some of these embodiments, it is useful to embed those frequencies that have been empirically found to be effective in achieving a desired outcome. These include combinations of frequencies, among which are those that are constituent frequencies of alpha, beta, delta, and theta waves. In such embodiments, the artificial-intelligence engine 92 has been trained to choose among the constituent frequencies based on a desired outcome.

[0100] In a composition that has multiple stems, it is desirable for the stems to be synchronized. Accordingly, a music model 94 that has been configured to handle multiple stems is preferably one that has been trained to synchronize all the stems, including those that add binaural beats and that embed particular frequencies. This promotes harmonic and rhythmic coherence of the composition as a whole.

[0101] In some embodiments, the artificial-intelligence engine 92 carries out further refinements to the composition during a post-processing phase. Examples of such refinements carried out during a post-processing phase include the refinement of embedded frequencies, for example by adjusting amplitudes thereof or by applying audio effects to ensure seamless integration of different stems into an overall composition. Such post-processing promotes the frequence embedding’s ability to carry out its intended therapeutic and / or psychoacoustic effects.

[0102] Examples of audio effects that the artificial-intelligence engine 92 applies during the post-processing phase include filtering with low-pass filters, high-passAttorney Docket No. 30263-008W01 filters, bandpass filters, reverberation, and echo, either within a stem or within the composition as a whole.

[0103] In still other embodiments, the artificial-intelligence engine 92 has been trained to adjust tempo and / or volume of a composition or particular tracks thereof and to do so based on preset criteria.

[0104] In those cases in which the subject’s target state is that of reducing drowsiness and attaining greater alertness, it is useful to modify the music’s structure by triggering a reward stimulus 78 when the measured feature set 30 indicates increasing power in the lower part of the subject’s beta- wave band and a concurrent drop in power present within the subject’s theta-wave band.

[0105] In some cases, the measured feature set 30 includes information concerning power levels obtained from different parts of the subject’s brain. This raises the possibility of the measured feature set 30 showing an asymmetry in the subject’s electroencephalogram. Such an asymmetry arises when the power levels for a given channel of the electroencephalogram vary significantly depending on where on the brain the transducer set 24 makes measurements. A common source of asymmetry arises when the same channel, e.g., the alpha-wave channel, is observed at opposite hemispheres of the brain. The resulting amplitude asymmetry is a useful indicator of the subject’s state. In some embodiments, when the subject 16 displays an asymmetry in the alpha- wave channel, the adaptation engine 54 attempts to promote a more relaxed state by increasing the music’s tonal brightness.

[0106] Still another example of using features within the electroencephalogram’s spectrum arises upon detecting that the frequency at which the peak power within the alpha-wave band occurs has begun to climb. In such cases, the adaptation engine 54 attempts to promote the subject’s alertness by increasing the music’s tempo or elevating its average pitch.

[0107] Transformations of musical structure in response to physiological signals 16 are sometimes carried out in a way that changes the subject’s perception of entropy in the music. In some embodiments, the adaptation engine 54 attempts to increase the perceived entropy by carrying out one or more of the foregoing transformations of musical structure with some degree of randomness or to decrease the amount of perceived entropy by carrying out one or more of the foregoing transformations with greater predictability.

[0108] The invention thus provides a neurofeedback system that carries out real-time adaptation of music in an effort to urge the subject 16 towards a target mental stateAttorney Docket No. 30263-008W01 and to stabilize the subject’s target mental state upon having achieved it. The realtime adaptation includes continuous transformations of the music’s musical structure that extend beyond mere static playback of the music.

[0109] The dynamically-adapted music thus functions as a vehicle for providing feedback that rewards the subject 16 when the subject 16 moves towards or maintains a target mental state. In effect, the distributed brain-training circuit and method apply principles of operant conditioning to draw the user into the target mental state.

[0110] In some embodiments, the adaptation engine 54 attempts to promote the efficacy of the operant conditioning method described herein by incorporating a scoring mechanism. A suitable scoring mechanism relies on composite indices of one or more features of the physiological signal 20.

[0111] In other embodiments, the adaptation engine 54 implements feature threshold- triggered transitions. An example of a threshold-triggered transition is one in which achievement of a particular score (i.e., “threshold”) causes a step-wise transition to a different operating mode and those in which the transition is instead a continuous modulation of the music’s musical structure.

[0112] In some embodiments, the personalization engine 52 has been trained based on the subject’s responses to music that has been dynamically adapted by the adaptation engine 54. Embodiments include those in which the personalization engine 52 trains itself in real time and those in which it has been trained before deployment.

[0113] The personalization engine 52 receives the physiological signal 21) from the feature signal 48, the target state information 46, and an event signal 98, which comes from an event sensor 100 that has been configured to provide the event signal 98 to the local circuit 12, which forwards it to the personalization engine 52.

[0114] The event signal 98 indicates the occurrence of the event, the nature of the event, and the time of its occurrence. A suitable way to indicate the nature of an event is to use a look-up table in which each event in a plurality of predefined events has been assigned a code, such as a particular integer. Information in an event signal 98 is useful for indicating a subject’s state because it has been discovered that a subject’s state is influenced by the occurrence of particular external events. As a result, the personalization engine 52 is able to combine information in the event signal 98 with that in the feature signal 48 to infer the subject’s state in real time.

[0115] Referring now to FIG. 3, the personalization engine 52 features an input layer 102 that includes a spectral analyzer 104. Among the spectral analyzer’s roles is that of continuously receiving the feature signal 48 in real time. The spectral analyzer 104Attorney Docket No. 30263-008W01 uses the feature signal 48 and its complex conjugate to evaluate the magnitude of physiological signal 20 as a function of that signal’s constituent frequencies. The result is a power spectrum that shows the first signal’s power spectrum.

[0116] A suitable implementation of the spectral analyzer 104 is one that comprises a multi-channel integrator 106 having plural integration circuits 108.

[0117] Each integration circuit 108 evaluates an integral of the power spectrum from a lower frequency limit to an upper frequency limit. The frequencies between the lower frequency limit and the upper frequency limit define a “channel.” The resulting integral, when normalized by the difference between the lower frequency limit and the upper frequency limit, yields a value that is indicative of the average power in that channel. Because it also receives the power as a function of frequency in each channel, each integration circuit 108 is able to also determine the frequency within that channel that carries the most power. This particular frequency will be referred to herein as the “peak frequency” for that channel.

[0118] Among the embodiments of the multi-channel integrator 106 are those that have two or more integration circuits 108 that are configured to carry out integration of the power spectrum across any two or more of: a delta-wave channel, a theta-wave channel, an alpha-wave channel, a beta-wave channel, and a gamma-wave channel.

[0119] Also among the embodiments of the multi-channel integrator 106 are those that integrate the power spectrum in any two or more of a channel that spans a range between a half cycle per second and four cycles per second, a channel that spans arrange between four and eight cycles per second, a channel that spans a range between eight and thirteen cycles per second, a channel that spans a range between thirteen and thirty cycles per second, and a channel that spans a range between thirty and a hundred cycles per second.

[0120] The output of the spectral analyzer 104 is thus a vector output in which each element of the vector represents the output of one of the multi-channel integrator’ s constituent integrator circuits 108. Thus, in the case of a multi-channel integrator 106 that decomposes the physiological signal 20 into delta waves, theta waves, alpha waves, beta waves, and gamma waves, the spectral analyzer’s output is a fivedimensional vector in which each element of the vector provides information about a corresponding one of the five channels. In some embodiments, these elements include both a real part and an imaginary part, which can then be used to determine a magnitude and phase for each channel.Attorney Docket No. 30263-008W01

[0121] The spectral analyzer’s output is provided to a channel -pair selector 110. The channel-pair selector 110 selects two of the channels and provides them to a divider circuit 112. The divider circuit’s output indicates a ratio of power present within a first channel to power present within a second channel. In some embodiments, the adaptation engine 54 uses power ratios between pairs of channels as a basis for inferring the subject’s state.

[0122] Within each of the channels, the distribution of power as a function of frequency need not be constant. It is often the case that within a particular channel, there exists a frequency at which the highest power level occurs. In some embodiments, the adaptation engine 54 uses the location of this frequency as a basis for inferring the subject’s state.

[0123] Accordingly, in some embodiments, the spectral analyzer 100 provides its outputs to a multi-channel peak detector 116 that includes peak detectors 116, each of which corresponds to one of the channels. Each peak detector 116 scans the power at each frequency within its assigned channel and identifies that frequency within that band that has the highest power level. In such embodiments, the adaptation engine 54 relies at least in part on the frequency of, for example, the alpha wave having the highest power level, as a basis for inferring the subject’s state.

[0124] In some embodiments, the spectral analyzer’s output is provided to a phase detector 118, which then compares the phase difference between any pair of the spectral analyzer’s outputs. This provides a way to determine a phase-locking value (“PLV”) that is indicative of coherence between the signals at a particular pair of outputs. In such embodiments, the adaptation engine 54 relies at least in part on this coherence to infer the subject’s state.

[0125] Certain other embodiments include those in which the personalization engine 52 infers the subject’s state based at least in part on a relationship between the feature signal 48 and the event signal 98 provided by the event sensor 100. The event signal 98 includes a time stamp to indicate when the event occurred. Based on the foregoing information, the adaptation engine 54 identifies event-related potentials, which are then relied upon in the course of inferring the subject’s mental state.

[0126] In still other embodiments, the personalization engine 52 infers the subject’s mental state based on the manner in which power is distributed across frequency in the physiological signal 20. Of particular interest is the complexity and unpredictability of the manner in which the input signal is spread across the frequency bands. To provide this information, the spectral analyzer provides its output to aAttorney Docket No. 30263-008W01 spectral-entropy analyzer 120, the output of which is a metric that is indicative of entropy and / or complexity in the input signal.

[0127] The personalization engine 52 also includes a state-detection layer 122 that receives the foregoing indicators of mental state. The state-detection layer 122 incorporates a classifier 124 that has been trained, or that has trained itself, to identify mental state based on those indicators. Examples of a classifier 124 include a support vector machine and a neural network, including a deep neural network.

[0128] Among the operations carried out by the state-detection layer 122 is that of detecting any statistically significant or algorithmically meaningful changes in features that are present in the physiological signal and doing so in real time. Such changes are indicative of spontaneous or induced shifts in mental state. Examples of such changes include changes in the subject’s level of attention, changes in the subject’s ability to focus, changes in the subject’s state of relaxation, changes in the subject’s state of calmness, changes in the subject’s level of arousal, and changes in the subject’s alertness. Other examples of spontaneous or induced shifts that are detectable based on statistically significant or algorithmically meaningful changes in features that are present in the physiological signal include changes wrought by external stimuli, such as a change in the cognitive load experienced by the subject 16.

[0129] Changes in features of the physiological signal 20, and in particular, those changes that carry either statistical or algorithmic significance, also indicate, in real time, spontaneous or induced shifts in the manner in which the subject 16 is carrying out a mental process, such as whether or not the subject 16 has transitioned into or out of a flow state or whether the subject 16 is in a meditative or quasi-meditative state, and if so, the extent to which the subject 16 is in that state, i.e., the subject’s meditative depth.

[0130] In many cases, a subject 16 experiences a spontaneous or induced shift in emotional state. Such a shift can be masked through the subject’s conscious effort. Indeed, the ability to consciously conceal such a shift is often regarded as a talent to be cultivated for the sake of promoting socialization of the subject 16. As a result, shifts in emotional state, for example changes in emotional valence, are sometimes imperceptible to an outside observer.

[0131] On the other hand, such shifts in emotional valence are readily manifested by changes in the features in the physiological signal 20, and particularly those changes that are of statistical or algorithmic significance. Accordingly, in some embodiments, the state-detection layer 122 is configured to detect such features in the physiological signal 20.Attorney Docket No. 30263-008W01

[0132] In some cases, statistically significant or algorithmically meaningful changes in features that are present in the physiological signal 20 indicate, instead of a shift, a readiness to make such a shift, i.e., a readiness to transition between two states of consciousness or to transition between a conscious state and an unconsciousness, such as a readiness to fall asleep or to awaken. Accordingly, in some embodiments, the state-detection layer 122 is configured to detect, within the physiological signal 20, features that are indicative of a shift into a state of being ready to make another shift.

[0133] The brain- training circuit 10 thus forms a distributed closed-loop feedback circuit in which the personalization engine 52 and the adaptation engine 54 cooperate to expose the subject 16 to a combined stimulus 82 in an effort to guide the subject’s brain waves towards having features that are present in the target feature-set 60 through exposure to the combined stimulus 82, with the combined stimulus 82 being adapted in real time in an effort to guide the received feature set towards the target feature- set 60.

[0134] FIG. 4 shows a method 124 in which the brain-training circuit 10 uses both real-time brain signals and predetermined state criteria to guide real-time changes in that portion of music that is a constituent of the conditioning- audio stimulus 80 so as to both guide the subject’s brain into the target state and condition the subject’s brain to learn how to maintain that target state. The brain signals are captured by the headset 12. The predetermined state criteria are embodied by the target feature- set 60.

[0135] The method 124 begins with receiving a measured feature-set 58 derived from the brain signal (step 126) and inspecting that measured feature-set 58 to see if it conforms to a target feature-set 60 indicative of a desired brain state (step 128). If so, the reward signal 84 causes the reward switch 86 to add the reward stimulus 78 to the conditioning-audio stimulus 80 (step 130). Otherwise, the conditioning-audio stimulus 80 consists of only the base-audio stimulus 74 (step 132). In either case, the result is an updated conditioning-audio stimulus 80 that has been modified based on real-time changes to a received electroencephalogram from the subject 16 (step 134). The resulting conditioning- audio stimulus 80 is then provided to the subject 16 (step 136). As a result of exposure to the conditioning-audio stimulus 80, changes occur to the subject’s brain state (step 138). These changes are manifested as an update to the measured feature-set 58 (step 140).

[0136] Referring now to FIG. 5, the base-audio stimulus 74 itself can be viewed as a superposition of frequency components weighted by, in general, complex values. The adaptation engine 54 would therefore be able to also vary the conditioning- audio stimulus 80 by modifying this superposition of the music’s components. In suchAttorney Docket No. 30263-008W01 cases, the reward signal 84 triggers operation of an audio-modification circuit 126. After having received the base-audio stimulus 74 from the audio generator 68, the audio-modification circuit 142 modifies it to form the conditioning-audio stimulus 80.

[0137] In some cases, the components of the base-audio stimulus 74 form an orthogonal basis of a function space. For example, the components can be complex exponentials. In such cases, the audio-modification circuit 142 adaptively varies the combined stimulus 82 to suppress or enhance certain frequencies of the base-audio stimulus 74 in an attempt to drive the brain waves towards having the target feature- set 60.

[0138] In other cases, the components of the base-audio stimulus 74 do not form an orthogonal basis of the function space. For example, a first component could be the function that, when played by itself, contains the sounds made by the string section and a second component could be the function that, when played by itself, sounds the rest of the orchestra minus the string section from the first component. This granularity of components can be further increased. For example, the components may include a function that contains the sound played by a particular violin.

[0139] In either case, the audio-modification circuit 142 can individually weight the components whose superposition forms the base-audio stimulus 74, for example by a complex number, so as to modify the amplitude of that component and its phase relative to other components in an attempt to synthesize a conditioning-audio stimulus 80 to drive the features obtained from the feature signal 48 towards the target feature- set 60. In effect, this generalizes the concept of the reward stimulus 78 from being a particular track that is simply added to the base-audio stimulus 74 to being an operator that operates on the base-audio stimulus 74 in a manner that is more complex than simply adding a track.

[0140] It should be noted that the act of modifying an existing musical composition by operating on it, for example by assigning weights to its components, effectively amounts to creating either a derivative work or a new composition, depending on the extent of the modification. As a result, the audio-modification circuit 142 can be viewed as an adaptive music arranger that adaptively generates a base-audio stimulus 74 or improvises off an existing base-audio stimulus 74 in an effort to condition neurons in the subject’s brain to achieve a desired state, the desired state having been defined by the target features.

[0141] In still other embodiments, either the audio or video stimuli 46, 48 of the combined stimulus 82 is adaptively modified based on changes in brain state or in neural activity. Examples include causing the base-audio stimulus 74 to pause,Attorney Docket No. 30263-008W01 changing the overall volume of the base-audio stimulus 74 as a whole or on a component-by-component basis, or changing the perceived source of the audio constituent, for example by varying the relative volumes heard on either loudspeaker of the wearable platform ^>2.

[0142] In some embodiments, the audio-modification circuit 142 causes the conditioning- audio stimulus 8 to comprise adaptive music that responds in real time to real-time changes in the subject’s physiology and, in particular, on activity within the subject’s brain. In effect, the conditioning-audio stimulus 80 comprises music that has been modified in such a way that various aspects of music will change in real time as in response to the subject’s brain activity.

[0143] A variety of changes are contemplated as the conditioning- audio stimulus 80 adapts to different conditions. In some examples, pre-recorded tracks or tracks generated in real time by the artificial-intelligence engine 84 are added or removed from the conditioning- audio stimulus 80 in response to brain-related changes. In some cases, these tracks correspond to particular musical instruments or groups thereof. In other examples, the volume of the conditioning-audio stimulus 8 changes in response to brain-related changes in the subject 16.

[0144] Examples of the foregoing include embodiments in which the audiomodification circuit 142 changes the volume of a particular component of the baseaudio stimulus 74 in response to brain-related changes in the subject 16. This would include the limiting case in which the audio-modification circuit 142 reduces the volume to zero, as a result of which the component disappears. Among these are embodiments in which the component whose volume is changed is a musical track.

[0145] The conditioning-audio stimulus 80 includes those components that are relied upon to achieve operant conditioning and those that are not. The latter components are inactive components that serve primarily as a background upon which the former components, i.e., the active components, can be overlaid. Since these inactive components are not relied upon for operant conditioning, they can be altered without compromising the effectiveness of operant conditioning.

[0146] In an effort to simultaneously achieve operant conditioning while avoiding repetition, certain embodiments include changing the inactive components while preserving the active components. This manifests as changing selected musical track volumes randomly or algorithmically. This ensures that the subject 16 perceives novelty in the music while preserving those components of the conditioning-audio stimulus 80, i.e., the active components, that are responsible for operant conditioning.Attorney Docket No. 30263-008W01Accordingly, the passive components serve primarily a cosmetic role to avoid what otherwise might be perceived as annoying repetition of the active components.

[0147] In those embodiments that have two independent audio channels, one for each ear, the change to the conditioning-audio stimulus 80 takes the form of a change to the properties of the first and second audio channels independently of each other. For example, the volume in one audio channel may increase relative to that in the other channel in response to detection of brain-related changes in the subject 16. This can take place gradually and unobtrusively so as to cause the illusion of movement in the source of the sound.

[0148] In some embodiments, such changes to the conditioning-audio stimulus 80 occur as a result of real-time changes in the subject’s brain activity. These real-time changes include those indicative of having achieved a success threshold, those indicative of an achievement associated with passage of time or completion of an event, based on having achieved some criterion for a specified time, based on having achieved a sufficiently high level of some criterion for a sufficiently long time. Also contemplated are embodiments in which a variable schedule of reinforcement is used, in which case the changes to the conditioning-audio stimulus 80 will occur at random times according to some specified probability distribution.

[0149] Achievement of a success threshold is manifested by having achieved a predetermined threshold value based on combinations of features in the measured feature set 30, such as total amplitude within a band and other relationships between bands. Examples of such bands include the delta, theta, SMR, mid beta, high beta, and gamma bands.

[0150] Embodiments further include those in which the conditioning-audio stimulus 80 comprises music that has been psycho-acoustically manipulated to help achieve particular predetermined target feature-sets 32 such as those associated with relaxation, meditation, and focus. Such manipulations include manipulations to the musical structure, such as manipulations of tempo, rhythm, instrumentation, melodic patterns, harmonics, instrumentation, frequency emphasis, and orchestration. These manipulations have been selected by the adaptation engine 54 to promote a particular state-of-mind corresponding to a target feature-set 60.

[0151] Also among the embodiments are those in which the conditioning- audio stimulus 80 includes a pair of tones having almost the same frequency so as to cause perception of binaural beating, or a tone that is turned on and off at some regular rate, i.e., an iso-chronic tone. Also among the embodiments are those in which the conditioning-audio stimulus 8 includes one or more of: strategic spatial audioAttorney Docket No. 30263-008W01 changes, enhancement of one or more particular frequencies, and the embedding of one or more frequencies.

[0152] A variety of sources are available for music associated with the base-audio stimulus 74. In some embodiments, the music has been specially composed for the occasion. In others, the music comes from a third-party music engine that has been pre-categorized by the artificial-intelligence engine 84 and further processed to add appropriate conditioning elements as described above. In still other embodiments, an artificial-intelligence engine 84 incorporated into the audio generator 68 has been trained to provide music having features that promote a transition into the target state.

[0153] In either case, the result is music that has been designed using neuroscientific and psycho-acoustic methods to promote achievement of particular states. Such music design includes manipulation of one or more musical and psycho-acoustic variables that are part of the music’s musical structure. Examples include including tempo, rhythm, tones, including overall frequency balance and / or emphasis on lower or higher frequencies, such as bass and treble frequencies, timbre, musical texture, resonance, entrainment, which promotes a temporal locking of various physiological phenomena, such as motor activity, respiration, heart rate, and brain activity, with an external periodic signal, and overtones, which are used to reinforce perception of a fundamental frequency.

[0154] In some embodiments, the reward stimulus 78 comprises additional tracks that have been designed and composed strategically and specifically to be added to the main music track. Such a reward stimulus 78 is activated and heard in real-time as an additional layer of sound in addition to the base-audio stimulus 74 only when the subject 16 has attained a pre-determined brain state. The subject’s perception of this reward stimulus 78 provides feedback to the subject 16. This feedback indicates, to the subject 1 , attainment of the target feature- set 60. In some embodiments, the reward stimulus 78 is a subtle one. An example of such a reward stimulus 78 is one that adds harmonics or overtones to the base-audio stimulus 74 so as to cause the subject 16 to perceive a greater fullness to the sound. In other embodiments, the reward stimulus 78 also includes additional frequencies that have been strategically chosen to induce a particular brain state.

[0155] Because the brain-training circuit 10 has access to real-time electroencephalograms, it is able to learn the subject’s response to particular stimuli. This allowed implementation of a machine-learning approach to enable the reward stimulus 78 to be fine-tuned to specific and personalized features of the subject’s physiological signals 16, such as the subject’s brain signature. As a result, it isAttorney Docket No. 30263-008W01 possible to cause the reward stimulus 78 to adaptively change based on a subject’s moment-to-moment success in moving towards the desired brain state.

[0156] In some embodiments, the alpha-peak frequency is calculated as a basis for automatically choosing one or more aspects of the reward stimulus 78. This process begins with monitoring certain features of a baseline electroencephalogram prior to a training session and calculating the alpha-peak frequency. Additional features of the reward stimulus 78 that match a particular base-audio stimulus 74 are generated or pre-composed with embedded frequencies to promote triggering of particular brain states. These embedded frequencies include overtones, resonant frequencies, and harmonics that harmonize with the musical key of the base-audio stimulus 74.

[0157] In the discussion that follows, which concerns FIGS. 6-8, there will be references to certain reference numerals that have also been used in connection with FIGS. 1-5. To the extent a reference numeral is used to identify a component in FIGS. 6-8 that differs from the component that that that reference numeral identifies in FIGS. 1-5, that reference numeral should be understood in the context of FIGS. 6-8 only. Stated in terms familiar from the field of computer science, the “scope” of reference numerals in FIGS. 6-8 does not extend to FIGS. 1-5 and vice versa.

[0158] FIG. 6 shows distributed feedback-circuitry 10 for taking advantage of neuroplasticity to condition neurons in a subject’s brain. The distributed feedbackcircuitry 10 conditions neurons in the subject’s brain so that they cooperate to cause a transition into a desired mental state and to thereafter cause the neurons to remain in that mental state. Examples of mental states include a relaxed state, a stress-reducing state, a state conducive to sleep, a state conducive to enhancement of focus, and a state conducive to greater attention. Other examples include a state of mindfulness, an energetic state, a happy state, and a state of flow.

[0159] The distributed feedback-circuitry 10 relies in part on a wearable platform 12 that carries a signal transducer 14.

[0160] FIG. 6 shows a particular embodiment in which the wearable platform 12 comprises a headset and the signal transducer 14 comprises an electroencephalogram that is built into a headset. Other examples of signal transducers 14 include a heartrate monitor, which provides information on heart rate, changes in heart rate, and heart-rate variability, a temperature monitor for measuring body temperature, including peripheral body temperature, a galvanometer that measures galvanic skin response and changes thereof, an electromyograph that monitors changes in muscle tension, and a spirometer that measures respiration rate or respiration volume.Attorney Docket No. 30263-008W01

[0161] The ensemble of transducers 14 outputs a first measurement signal 16 that includes information concerning the real-time physiological signals measured by each of the transducers 14. The first measurement signal 16 is then transmitted in digitally- encoded form to local circuitry 18 that executes a training application 20.

[0162] Examples of local circuitry 18 include circuitry embedded in various portable and / or wearable devices, such as a smartphone, a tablet, smart jewelry, a smart watch, and a laptop. Other examples of local circuitry 18 include non-portable devices, such as a personal computer.

[0163] In operation, the training application 20 will have received, from the subject, instructions indicative of a target mental state of that subject. The distributed feedback-circuitry 10 then attempts to urge the subject’s mental state towards this target mental state based at least in part on information from the first measurement signal 16.

[0164] The training application 20 displays information from the first measurement signal 16 in real time on the local circuitry 18. In addition, the training application 20 causes the local circuitry 18 to forward a second measurement signal 22 to biofeedback circuitry 24.

[0165] In some embodiments, the biofeedback circuitry 24 executes on hardware 25 in the cloud. In such embodiments, the biofeedback circuitry 24 is shared by plural instances of the local circuitry 18 via a network.

[0166] In other embodiments, the biofeedback circuitry 24 is separate from the local circuitry 18 but used only by that local circuitry 18. An example of such an embodiment is one in which the hardware 25 that executes the biofeedback circuitry comprises a robot, and in particular, a robot having access to artificial intelligence circuitry, which is either built into the robot itself or accessible remotely. An example of such a robot is one manufactured by HANSEN ROBOTICS that has been given the tradename “SOPHIA.” In still other embodiments, the hardware 25 that executes the biofeedback circuitry 24 is distributed between the robot and a cloud server.

[0167] The second measurement signal 22 incorporates information derived from the first measurement signal 16. In those cases in which the signal transducers comprise an electroencephalogram, examples of such information include information concerning changes in amplitude or power of any frequency or combination of frequencies present in an electroencephalogram provided by the electroencephalograph, including the presence of frequencies at ten hertz and forty hertz. Other examples of such information include information concerning changes inAttorney Docket No. 30263-008W01 amplitude or power or ratios of alpha, beta, gamma, delta, and theta waves from the foregoing encephalogram.

[0168] Still other examples of information derived from the first measurement signal 16 include information concerning changes in coherence or phase between various components of the foregoing electroencephalogram; information concerning changes in ratios of bands or frequencies in the foregoing electroencephalogram; and information concerning changes in in left / right symmetry or in the symmetry at two or more locations from the foregoing encephalogram.

[0169] In addition, the training application 20 causes the local circuitry 18 to transmit target-state information 26 to the biofeedback circuitry 24. This target-state information 26 comprises information indicative of what sort of neural conditioning the subject wishes to achieve.

[0170] The biofeedback circuitry 24 includes a feature-extraction circuit 28 that carries out feature extraction on the second measurement signal 22 to obtain a measured feature- set 30 for the subject. Based on the target- state information 26, the feature-extraction circuit 28 defines a target feature-set 32. It then provides both the target feature-set 32 and the measured feature-set 30 to a controller 34.

[0171] The controller 34 uses the target feature-set 32 and the measured feature-set 30 as a basis for providing a control signal 36 to a conditioner 38. The conditioner 38 comprises an audio generator 40, a video generator 42, and a reward generator 44. The audio generator 42 provides a base-audio stimulus 46; the video generator 42 provides a video stimulus 48; and the reward generator 44 provides a reward stimulus 50. The base-audio stimulus 46 and the occasional reward stimulus 50 combine to form a conditioning- audio stimulus 60. The conditional-audio stimulus 60 is then combined with the video stimulus 58 to form a conditioning stimulus 52. It is this conditioning stimulus 52 that is ultimately provided to the headset 12. This conditioning stimulus 52 provides positive or negative reinforcement to the subject in an attempt to condition the subject to reach the target feature-set 32.

[0172] The existence of positive or negative reinforcement in the conditioning stimulus 52 arises as a result of the controller 34 providing a reward signal 54 to control a reward switch 56. Depending on the state of the reward switch 56, a first adder 58 either combines the reward stimulus 50 with the base-audio stimulus 46 and removes the reward stimulus 50 from the base-audio stimulus 46. This controls whether the resulting conditioning-audio stimulus 60 provides positive reinforcement or negative reinforcement to the subject. A second adder 62 receives the conditioning-Attorney Docket No. 30263-008W01 audio stimulus 60 and combines it with the video stimulus 48 to form the conditioning stimulus 52.

[0173] The reward signal 54 provided by the controller 34 thereby controls whether or not the conditioning stimulus 52 will comprise a superposition of the audio stimulus 60 and the reward stimulus 50 or just the audio stimulus 60 by itself. By controlling whether the reward stimulus 50 is present or absent based on the difference between the target feature-set 32 and the measured feature-set 30, the controller 34 carries out the conditioning in an effort to drive the measured-feature set 30 closer to the target feature-set 32.

[0174] The reward stimulus 50 is made to appear or disappear or is otherwise altered in response to the progress being made towards arriving at the target feature-set 32. In some embodiments, the reward stimulus 50 is a single tone whereas in others it is a weighted combination of frequencies having complex-valued weighting coefficients.

[0175] The subject’s neurons are thus exposed to the conditioning stimulus 52 via the auditory pathway. This enables the conditioning process to be carried out. The controller 34 tailors the conditioning stimulus 52 to cause the measured feature-set 30 to transition into the target feature-set 32. It does so by reducing a difference between the measured feature-set 30, which is arriving at the controller 34 in essentially real time, and the target feature-set 32, which was provided by the training application 20 at the beginning of the conditioning process. This difference is reduced at least in part by the reward signal 54, which selectively adds or subtracts the reward stimulus 50 from the conditioning stimulus 52 as well as by the selection of content the reward stimulus 50 and the content of both the audio stimulus 60 and the video stimulus 48 so as to condition the subject’s relevant neurons so that they achieve and maintain a desired brain state.

[0176] The biofeedback circuitry 24 transmits the conditioning stimulus 52 to the training application 20. Upon receipt of the conditioning stimulus 52, the training application 20 separates the video stimulus 48 from the conditioning-audio stimulus 60 and displays the video stimulus 48 on the local circuitry 18. The training application 20 then sends the conditioning-audio stimulus 60 to the headset 12 to be listened to by the subject. As a result, the subject’s neurons are exposed to the conditioning stimulus 52 using two different sensory pathways.

[0177] Like the conditioning-audio stimulus 60, the visual stimulus 48 reflects the real-time electroencephalogram measurements obtained from the subject. Accordingly, the details of the visual stimulus 48 change in response to changes in those electroencephalogram measurements. Preferably, the video stimulus 48 and theAttorney Docket No. 30263-008W01 conditioning-audio stimulus 60 are synchronized. As a result, when the conditioningaudio stimulus 60 is changed, the video stimulus 48 sustains a corresponding change.

[0178] The training application 20 continues to receive measurement signals 60 from the electroencephalograph 14 as the subject is exposed to the conditioning stimulus 52. These are transmitted to the biofeedback circuitry 24 to serve as a basis for feedback control over the neural conditioning process.

[0179] The biofeedback circuitry 24 carries out further feature extraction on the second measurement signal 36. The resulting updated measured feature-sets 30 provide a basis for evaluating the effect of the conditioning stimulus 52 and, in particular, the progress made towards driving the measured feature-set 30 towards the target feature-set 32. In response to the assessment of such progress, the controller 34 causes the conditioner 38 to formulates a revised conditioning-stimulus 52. It then transmits the revised conditioning-stimulus 52 back to the training application 20 so that the neurons to be conditioned can be exposed to them via the subject’s sensory pathways.

[0180] The distributed feedback-circuitry 10 thus forms a distributed closed- loop feedback circuit that exposes the subject to a conditioning stimulus 52 in an effort to guide the subject’s brain waves towards having the target feature set 32 through exposure to conditioning stimulus 52, with the conditioning stimulus 52 being adapted periodically in an effort to guide the received feature set towards the target feature set 32.

[0181] Moreover, the biofeedback circuitry 24 is available for concurrent use by additional subjects 64 who interact with the biofeedback circuitry 24 in the same manner using corresponding headsets 12 and local circuitry 18.

[0182] FIG. 7 shows a method 66 carried out by the distributed feedback-circuitry 10 using real-time brain signals, which are captured by the headset 12, and predetermined brain state criteria, as embodied by the target feature-set 32, to guide real-time changes in music that is a constituent of the conditioning-audio stimulus 60 so as to guide the subject’s brain into a desired state and to condition the subject’s brain to learn how to maintain that state.

[0183] The method 66 begins with receipt of a measured feature-set 30 derived from the brain signal (step 68) and an inspection to see if that feature set conforms to a target feature-set 32 indicative of a desired brain state (step 70). If so, the reward signal 54 causes the reward switch 56 to add the reward stimulus 50 to the conditioning-audio stimulus 60 (step 72). Otherwise, the conditioning- audio stimulusAttorney Docket No. 30263-008W0160 consists of only the base-audio stimulus 46 (step 74). In either case, the result is an updated conditioning- audio stimulus 60 that has been modified based on real-time changes to a received electroencephalogram from the subject (step 76). The resulting conditioning- audio stimulus 60 is then provided to the subject (step 78). As a result of exposure to the conditioning- audio stimulus 60, changes occur to the subject’s brain state (step 80). These changes are manifested as an update to the measured feature-set 30 (step 82).

[0184] Referring now to FIG. 8, the base-audio stimulus 46 itself can be viewed as a superposition of frequency components weighted by, in general, complex values. The biofeedback circuitry 24 would therefore be able to also vary the conditioning- audio stimulus 60 by modifying this superposition of the music’s components. In such cases, the reward signal 54 triggers operation of an audio-modification circuit 85 that receives the base-audio stimulus 46 from the audio generator 40 and modifies it directly to form the conditioning- audio stimulus 60.

[0185] It should be apparent that FIG. 8 is a generalization of FIG. 6. The combination of the reward generator 44, the reward switch 56, and the first adder 58 as shown in FIG. 6 can readily be viewed as a particular implementation of the more general audio-modification circuit 85 in FIG. 8.

[0186] In some cases, the components of the base-audio stimulus 46 form an orthogonal basis of a function space. For example, the components can be complex exponentials. In such cases, the audio-modification circuit 85 adaptively varies the conditioning stimulus 52 to suppress or enhance certain frequencies of the base-audio stimulus 46 in an attempt to drive the brain waves to have the target feature-set 32.

[0187] In other cases, the components of the base-audio stimulus 46 do not form an orthogonal basis of the function space. For example, a first component could be the function that, when played by itself, contains the sounds made by the string section and a second component could be the function that, when played by itself, sounds the rest of the orchestra minus the string section from the first component. This granularity of components can be further increased. For example, the components may include a function that contains the sound played by a particular violin.

[0188] In either case, the audio-modification circuit 85 can individually weight the components whose superposition forms the base-audio stimulus 46, for example by a complex number, so as to modify the amplitude of that component and its phase relative to other components in an attempt to synthesize a conditioning-audio stimulus 60 to drive the features obtained from the second measurement signal 36 towards the target feature-set 32. In effect, this generalizes the concept of the reward stimulus 50Attorney Docket No. 30263-008W01 from being a particular track that is simply added to the base-audio stimulus 46 to being an operator that operates on the base-audio stimulus 46 in a manner that is more complex than simply adding a track.

[0189] It should be noted that the act of modifying an existing musical composition by operating on it, for example by assigning weights to its components, effectively amounts to creating either a derivative work or a new composition, depending on the extent of the modification. As a result, the audio-modification circuit 85 can be viewed as adaptively composing a base-audio stimulus 46 or improvising off an existing base- audio stimulus 46 in an effort to condition neurons in the subject’s brain to achieve a desired state, the desired state having been defined by the target features.

[0190] In still other embodiments, either the audio or video stimuli 46, 48 of the conditioning stimulus 52 is adaptively modified based on changes in brain state or in neural activity. Examples include causing the base-audio stimulus 46 to pause, changing the overall volume of the base-audio stimulus 46 as a whole or on a component-by-component basis, or changing the perceived source of the audio constituent, for example by varying the relative volumes heard on either loudspeaker of the headset 12.

[0191] In some embodiments, the audio-modification circuit 85 causes the conditioning- audio stimulus 60 to comprises adaptive music that responds in real time to real-time changes in the subject’s physiology and, in particular, on activity within the subject’s brain. In effect, the conditioning-audio stimulus 60 comprises music that has been composed or modified in such a way that various aspects of music will change in real time as in response to the subject’s brain activity.

[0192] A variety of changes are contemplated as the conditioning- audio stimulus 60 adapts to different conditions. In some examples, pre-recorded tracks are added or removed from the conditioning- audio stimulus 60 in response to brain-related changes. In some cases, these tracks correspond to particular musical instruments or groups thereof. In other examples, the volume of the conditioning-audio stimulus 60 in response to brain-related changes in the subject.

[0193] Examples of the foregoing include embodiments in which the audiomodification circuit 85 changes the volume of a particular component of the baseaudio stimulus 46 in response to brain-related changes in the subject. This would include the limiting case in which the audio-modification circuit 85 reduces the volume to zero, as a result of which the component disappears. Among these are embodiments in which the component whose volume is changed is a musical track.Attorney Docket No. 30263-008W01

[0194] The conditioning-audio stimulus 60 includes those components that are relied upon to achieve operant conditioning and those that are not. The latter serve primarily as a background upon which the former can be overlaid. Since these latter components are not relied upon for operant conditioning, they can be altered without compromising the function of operant conditioning.

[0195] In an effort to simultaneously achieve operant conditioning while avoiding repetition, certain embodiments include changing the latter components while preserving the former. This manifests as changing selected musical track volumes randomly or algorithmically. This ensures that the subject perceives novelty in the music while preserving those components of the conditioning- audio stimulus 60 that are responsible for operant conditioning.

[0196] In those embodiments that have two independent audio channels, one for each ear, the change to the conditioning-audio stimulus 60 takes the form of a change to the properties of the first and second channels independently of each other. For example, the volume in one channel may increase relative to that in the other channel in response to detection of brain-related changes in the subject. This can take place gradually to cause the illusion of movement in the source of the sound.

[0197] In some embodiments, such changes to the conditioning-audio stimulus 60 occur as a result of real-time changes in the subject’s brain activity. These real-time changes include those indicative of having achieved a success threshold, those indicative of an achievement associated with passage of time or completion of an event, based on having achieved some criterion for a specified time, based on having achieved a sufficiently high level of some criterion for a sufficiently long time. Also contemplated are embodiments in which a variable schedule of reinforcement is used, in which case the changes to the conditioning-audio stimulus 60 will occur at random times according to some specified probability distribution.

[0198] Achievement of a success threshold is manifested by having achieved a predetermined threshold value based on combinations of features in the measured feature set 30, such as total amplitude within a band and other relationships between bands. Examples of such bands include the delta, theta, SMR, mid beta, high beta, and gamma bands.

[0199] Embodiments further include those in which the conditioning-audio stimulus 60 comprises music that has been composed and psycho-acoustically manipulated to help achieve particular predetermined target feature-sets 32 such as those associated with relaxation, meditation, and focus. Such manipulations include manipulations in tempo, rhythm, instrumentation, melodic patterns, harmonics, instrumentation,Attorney Docket No. 30263-008W01 frequency emphasis, and orchestration that have been designed to promote a particular state-of-mind corresponding to a target feature-set 32.

[0200] Also among the embodiments are those in which the conditioning- audio stimulus 60 includes a pair of tones having almost the same frequency so as to cause perception of binaural beating, or a tone that is turned on and off at some regular rate, i.e., iso-chronic tones. Also among the embodiments are those in which the conditioning-audio stimulus 60 includes strategic spatial audio changes the enhancement of one or more particular frequencies, or the embedding of one or more particular frequencies.

[0201] A variety of sources are available for music associated with the base-audio stimulus 46. In some embodiments, the music has been specially composed for the occasion. In others, the music comes from a third-party music engine that has been pre-categorized by the artificial-intelligence engine 84 and further processed to add appropriate conditioning elements as described above. In still other embodiments, an artificial-intelligence engine incorporated into the audio generator 40 has been trained to compose music having features that promote a transition into the target state.

[0202] In either case, the result is music that has been designed using neuroscientific and psycho-acoustic methods to promote achievement of particular mental states. Such music design includes manipulation of one or more musical and psycho-acoustic variable including tempo, rhythm, tones, including overall frequency balance and / or emphasis on lower or higher frequencies, such as bass and treble frequencies, timbre, musical texture, resonance, entrainment, which promotes a temporal locking of various physiological phenomena, such as motor activity, respiration, heart rate, and brain activity, with an external periodic signal, and overtones, which are used to reinforce perception of a fundamental frequency.

[0203] In some embodiments, the reward stimulus 50 comprises additional tracks that have been designed and composed strategically and specifically to be added to the main music track. Such a reward stimulus 50 is activated and heard in real-time as an additional layer of sound in addition to the base-audio stimulus 46 only when the subject has attained a pre-determined brain state. The subject’s perception of this reward stimulus 50 provides feedback to the subject. This feedback indicates, to the subject, attainment of the target feature-set 32. In some embodiments, the reward stimulus 50 is a subtle one. An example of such a reward stimulus 50 is one that adds harmonics or overtones to the base-audio stimulus 46 so as to cause the subject to perceive a greater fullness to the sound. In other embodiments, the reward stimulus 50Attorney Docket No. 30263-008W01 also includes additional frequencies that have been strategically chosen to induce a particular brain state.

[0204] Because the distributed feedback-circuitry 10 has access to real-time electroencephalograms, it is able to learn the subject’s response to particular stimuli. This allowed implementation of a machine-learning approach to enable the reward stimulus 50 to be fine-tuned to specific and personalized features of the subject’s physiological signals, such as the subject’s brain signature. As a result, it is possible to cause the reward stimulus 50 to adaptively change based on a subject’s moment-to- moment success in moving towards the desired brain state.

[0205] In some embodiments, the alpha-peak frequency is calculated as a basis for automatically choosing one or more aspects of the reward stimulus 50. This process begins with monitoring certain features of a baseline electroencephalogram prior to a training session and calculating the alpha-peak frequency. Additional features of the reward stimulus 50 that match a particular base-audio stimulus 46 are generated or pre-composed with embedded frequencies to promote triggering of particular brain states. These embedded frequencies include overtones, resonant frequencies, and harmonics that harmonize with the musical key of the base-audio stimulus 46.

[0206] As noted above, in some embodiments, the audio generator 40 features an artificial-intelligence engine 84. The artificial-intelligence engine 84 is one that has been trained to select characteristics that urge the subject to achieve and / or maintain a target mental state as specified by the target-state information 26. In some cases, these compositions mimic human musical composition.

[0207] A useful artificial-intelligence engine 84 is one that relies on a music model 86. A music model 86 is a generative model that has been trained using a music dataset 88 that comprises music. A suitable implementation of a music model 86 is a neural network.

[0208] As a result of its having been trained, the music model 86 learns musical patterns, structures, and styles. In some embodiments, the music dataset 88 includes examples of melodies incorporating subtle frequency modulations or binaural beats.

[0209] It has been observed that certain harmonic progressions and certain embedded frequencies promote therapeutic or psychoacoustic effects. Accordingly, in some embodiments, the music model 86 is one that has been trained using a music dataset 88 in which the music incorporates such harmonic progressions and / or embedded frequencies. The music model 86 incorporates these as chords and / or accompaniment in the generated music. In some cases, the music model 86 generates harmonicallyAttorney Docket No. 30263-008W01 embedded chords that concurrently complement the generated melody while embedding specific frequencies that have been found to enhance cognitive or emotional responses based on discoveries of such responses through examination of brainwaves.

[0210] In some embodiments, the artificial-intelligence engine 84 embeds binaural beats into the melody. In other embodiments, the artificial-intelligence engine 84 generates music that promotes relaxation, focus, and other desired mental states by outputting the same note at slightly different pitches, i.e., slightly different frequencies, in different output channels.

[0211] By embedding binaural beats or other auxiliary frequency components within the music, the music model 86 generates compositions with additional layers of depth and complexity. These additional layers potentially enhance the subject’s auditory experience and, most importantly, influence the mental and / or emotional state of that subject.

[0212] In some cases, the music model 86 is one that has been trained to produce melodies that have been engineered to have enhanced emotion impact. The music model 86 does so by incorporating frequency modulation to create both harmonic and disharmonic effects in the generated music.

[0213] Particularly useful features of the music model 86 include its ability to adapt and improve over time, its fluency in the creation of music according to a variety of styles, and its application across various musical genres and industries.

[0214] Some embodiments also rely on having the artificial-intelligence engine 84 generate individual components of a musical piece, such as melody, harmony, bassline, drums, and other instruments, and to do so in a way that preserves each component as a separate audio track. These components can then be mixed and matched to assemble a generated composition.

[0215] In those embodiments in which the music model 86 has been trained to generate basslines, there exist embodiments in which the music model 86 has been trained on a music dataset 88 that incorporates low-frequency oscillations or sub-bass frequencies that have been selected to interact with binaural beats or other incorporated frequency components in an effort to promote a more immersive experience. Other embodiments feature music models 86 that have been trained specifically to generate basslines incorporate rhythmic low-frequency pulses that align with the beat to reinforce the desired mental state induced by binaural beats that are present in the music.Attorney Docket No. 30263-008W01

[0216] In other embodiments, the music model 86 has been trained on drum patterns, including those that comprise a superposition of pulse trains at different frequencies. Such superpositions represent drum tracks that include embedded pulse trains at specific frequencies. When subtle enough to avoid overpowering the primary rhythmic elements but not so subtle that they fill to contribute to the overall frequency embedding or superposition of pulse trains in a measurable way, these embedded pulse trains tend to have noticeable psychoacoustic consequences. In these cases, the music model 86 is further trained to ensure that the various drum patterns synchronize with any embedded binaural beats that may be present in a different stem or to ensure that these same drum patterns synchronize with frequency modulations that may be present in different stems.

[0217] To promote diversity of timbre, it is useful to train the artificial-intelligence engine 84 to provide stems that include additional instrumental parts, each with a different timbre, and to use similar frequency embedding to enhance the auditory effects of the composition as a whole as a result of having incorporated such stems. In some of these embodiments, it is useful to embed those frequencies that have been empirically found to be effective in achieving a desired outcome. These include combinations of frequencies, among which are those that are constituent frequencies of alpha, beta, delta, and theta waves. In such embodiments, the artificial-intelligence engine 84 has been trained to choose among the constituent frequencies based on a desired outcome.

[0218] In a composition that has multiple stems, it is desirable for the stems to be synchronized. Accordingly, a music model 86 that has been configured to handle multiple stems is preferably one that has been trained to synchronize all the stems, including those that add binaural beats and that embed particular frequencies. This promotes harmonic and rhythmic coherence of the composition as a whole.

[0219] In some embodiments, the artificial-intelligence engine 84 carries out further refinements to the composition during a post-processing phase. Examples of such refinements carried out during a post-processing phase include the refinement of embedded frequencies, for example by adjusting amplitudes thereof or by applying audio effects to ensure seamless integration of different stems into an overall composition. Such post-processing promotes the frequence embedding’s ability to carry out its intended therapeutic and / or psychoacoustic effects.

[0220] Examples of audio effects that the artificial-intelligence engine 84 applies during the post-processing phase include filtering with low-pass filters, high-passAttorney Docket No. 30263-008W01 filters, bandpass filters, reverberation, and echo, either within a stem or within the composition as a whole.

[0221] In still other embodiments, the artificial-intelligence engine 84 has been trained to adjust tempo and / or volume of a composition or particular tracks thereof and to do so based on preset criteria.

[0222] It is to be understood that the foregoing description is intended to illustrate and not to limit the scope of the invention, which is defined by the scope of the appended claims. Other embodiments are within the scope of the following claims.

Claims

Attorney Docket No. 30263-008W01What is claimed is:

1. A method comprising causing a local circuit and a remote circuit that are interconnected by a network to cooperate in using real-time music stimuli to urge a subject having a time-varying state to attain a target state, said remote circuit comprising an adaptation engine and a personalization engine, wherein both said target state and said time- varying state are mental states of said subject, wherein causing said local circuit and said remote circuit to cooperate in using said real-time music stimuli to urge said subject to attain said target state comprises: at said personalization engine of said remote circuit, receiving, from said local circuit, an electrical signal that carries information indicating said target state and receiving a first physiological-signal from said local circuit, said first physiological-signal being an electrical signal that carries information that represents said time-varying mental state at a first time; causing said adaptation engine to generate a first music stimulus based at least in part on said first physiological-signal and to transmit said first music stimulus to said local circuit for exposure thereof to said subject; at said personalization engine, receiving a second physiological-signal from said local circuit, said second physiological-signal being an electrical signal that carries information that represents said time-varying mental state at a second time, which follows said first time and determining, based at least in part on said second physiological-signal, that said subject’s mental state continues to differ from said target state; causing said adaptation engine to generate a second music stimulus based at least in part on said second physiological-signal and to transmit said second music stimulus to said local circuit for exposure thereof to said subject; at said personalization engine, receiving a third physiological-signal from said local circuit, said third physiological-signal being an electrical signal that carries information that represents said timevarying mental state at a third time, which follows said second time and determining, based at least in part on said third physiological-signal, that said subject’s mental state has moved toward from said target state.Attorney Docket No. 30263-008W012. The method of claim 1, wherein generating said first music stimulus comprises transforming musical structure of selected music based at least in part on features present in said first physiological-signal.

3. The method of claim 1, wherein generating said first music stimulus comprises using a generative model to generate said first music stimulus based at least in part on features present in said first physiological-signal.

4. The method of claim 1, wherein said first physiological-signal comprises an electroencephalogram having a first spectrum and a second spectrum, wherein said first spectrum consists of frequencies within a first band of frequencies, wherein said second spectrum consists of frequencies within a second band of frequencies, wherein said first and second band of frequencies are disjoint, wherein generating said first music stimulus is carried out based at least in part on a change in a ratio of power present in said first spectrum and power present in said second spectrum.

5. The method of claim 1, wherein said first physiological-signal comprises an electroencephalogram having a first spectrum and a second spectrum, wherein said first spectrum consists of frequencies within a first band of frequencies, wherein said second spectrum consists of frequencies within a second band of frequencies, wherein said first and second band of frequencies are disjoint, wherein generating said first music stimulus is carried out based at least in part on a change in power present in said first spectrum.

6. The method of claim 1, wherein receiving said first and second physiological- signals comprises receiving said signals at said remote circuit after said signals are transmitted to said remote circuit by said local circuit and wherein said local circuit is one of a plurality of local circuits, each of which receives physiological signals from a different subject and each of which provides said physiological signals to said remote circuit.

7. The method of claim 1, wherein generating said first music stimulus comprises generating said first stimulus based at least in part on an event signal that indicates occurrence of an external event that affects said time- varying state ofAttorney Docket No. 30263-008W01 said subject, whereby said first stimulus is generated based at least in part on both said event signal and on said first physiological-signal.

8. The method of claim 1, wherein said first physiological-signal comprises an electroencephalogram having plural spectra, each of which consists of frequencies within a corresponding band of frequencies that are not in any other band of frequencies, wherein generating said first music stimulus comprises selecting two of said spectra and determining said time-varying state at a particular based at least in part on a power ratio between said two of said spectra.

9. The method of claim 1, wherein said first physiological-signal comprises an electroencephalogram having plural spectra and wherein generating said first music stimulus comprises, for each spectrum among said spectra, identifying a frequency that consumes a largest share of power in said spectrum.

10. The method of claim 1, wherein said first physiological-signal comprises an electroencephalogram having plural spectra and wherein generating said first music stimulus comprises, for each spectrum among said spectra, using a spectral-entropy analyzer to output a level of entropy in said spectrum.

11. The method of claim 1, wherein determining from said second physiological- signal that said subject’s mental state continues to differ from said target state comprises providing information from said second physiological signal to a state-detection layer that incorporates a classifier for classifying said subject’s mental state, said classifier being selected from the group consisting of a neural network and a support vector machine.

12. The method of claim 1, wherein determining from said second physiological- signal that said subject’s mental state continues to differ from said target state comprises providing information from said second physiological signal to a state-detection layer and causing said state-detection layer to identify a shift in emotional valence of said time- varying state of said subject.Attorney Docket No. 30263-008W0113. The method of claim 1, wherein causing said adaptation engine to generate said second music stimulus based at least in part on said second physiological- signal comprises transforming musical structure of an existing composition.

14. The method of claim 1, wherein causing said adaptation engine to generate said second music stimulus based at least in part on said second physiological- signal comprises causing said adaptation engine to transform an existing musical composition’s rhythmic structure.

15. The method of claim 1, wherein causing said adaptation engine to generate said second music stimulus based at least in part on said second physiological- signal comprises causing said adaptation engine to transform harmonic structure of an existing musical composition.

16. The method of claim 1, wherein there exists an existing musical composition that comprises spectral content and wherein causing said adaptation engine to generate said second music stimulus based at least in part on said second physiological-signal comprises causing said adaptation engine to transform said spectral content of said existing musical composition.

17. The method of claim 1, wherein causing said adaptation engine to generate said second music stimulus based at least in part on said second physiological- signal comprises causing said adaptation engine to add a reward stimulus to an existing musical composition and to use said existing musical composition, with said reward stimulus having been added thereto, as said second music stimulus.

18. The method of claim 1, wherein causing said adaptation engine to generate a first music stimulus comprises transforming a rhythmic density of an existing musical composition.

19. The method of claim 1, wherein an existing musical composition has a spectral brightness and wherein causing said adaptation engine to generate a first music stimulus comprises transforming said spectral brightness of said existing musical composition.Attorney Docket No. 30263-008W0120. An apparatus for use by a plurality of subjects, each of said subjects having a time-varying state and being equipped with a corresponding local circuit from a plurality of local circuits, each of said time- varying states being a mental state of a corresponding one of said subjects from said plurality of subjects, wherein said apparatus comprises a remote circuit that is in communication with each of said local circuits from said plurality of local circuits over a network, wherein said remote circuit comprises: an adaptation engine and a personalization engine, wherein said personalization engine is configured: to receive physiological signals from each of said subjects, to receive target states from each of said subjects, each of said target states having been specified by a corresponding one of said subjects, and to provide, to said adaptation engine, information indicative of a current state of each of said subjects and wherein said adaptation engine is configured: to provide music stimuli for each of said subjects based at least on part on one or more physiological signals from each of said subjects and to provide said music stimuli to corresponding ones of said local circuit for exposure to corresponding ones of said subjects.

21. The apparatus of claim 20, further comprising distributed feedback-circuitry that uses operant conditioning to train a subject from said plurality of subjects to achieve a target state of consciousness, said distributed feedback-circuitry comprising biofeedback circuitry that comprises: a feature-extraction circuit that receives a real-time measurement signal that carries information concerning one or more physiological signals, wherein said feature-extraction generates a measured feature-set and a target feature-set therefrom, said target feature-set comprising features for achieving said target state of consciousness, a controller that is configured to cause transmission of a conditioning stimulus to be listened to by said subject, said conditioning stimulus being one that causes said measured feature-set to be driven towards said target feature-set, wherein said controller is configured to cause said conditioning stimulus to comprise a conditioning-audio stimulus that transitions between being a base-audio stimulus with no reward stimulus and being a base-audio stimulus that has been operated on to incorporate a reward stimulus therein, wherein said controller causes said transition based onAttorney Docket No. 30263-008W01 progress made in causing said measured feature-set to conform to said target feature-set, and wherein said controller causes an artificial-intelligence engine to generate said base-audio stimulus using a music model that has been trained on a music dataset to generate music that has features that cause said measured feature-set to be driven towards said target feature-set.

22. The apparatus of claim 21, wherein said music model comprises a neural network.

23. The apparatus of claim 21, wherein said music model is one that has been trained to compose music having characteristics that cause said measured feature-set to be driven towards said target feature-set.

24. The apparatus of claim 21, wherein said music model is one that has been trained to include binaural beats in music, said binaural beats having been selected to cause said measured feature-set to be driven towards said target feature-set.

25. The apparatus of claim 21, wherein said music model is one that has been trained to compose music that comprises frequency modulation, said frequency modulation having been selected to cause said measured feature-set to be driven towards said target feature-set.

26. The apparatus of claim 21, wherein said music model is one that has been trained to compose music that comprises an embedded frequency, said embedded frequency having been selected to cause said measured feature-set to be driven towards said target feature-set.

27. The apparatus of claim 21, wherein said music model is one that has been trained to add a baseline to music, said bassline having been selected to cause said measured feature-set to be driven towards said target feature-set.

28. The apparatus of claim 21, wherein said music model is one that has been trained to add a superposition of pulse trains to music, said pulse trains having been selected to cause said measured feature-set to be driven towards said target feature- set.Attorney Docket No. 30263-008W0129. The apparatus of claim 21, wherein said music model is one that has been trained to add a stem to music, said stem having been selected to cause said measured feature-set to be driven towards said target feature-set.

30. The apparatus of claim 21, wherein said music model is one that has been trained to add a stem to music and to synchronize said stem with said music, said stem having been selected to cause said measured feature-set to be driven towards said target feature-set.

31. The apparatus of claim 21, wherein said music model is one that has been trained to filter music through a passband, said passband having been selected to cause said measured feature-set to be driven towards said target feature-set.

32. The apparatus of claim 21, wherein said music model is one that has been trained to suppress a set of frequencies in said music, said set of frequencies comprising all frequencies above a first frequency, said first frequency having been selected to cause said measured feature-set to be driven towards said target feature- set.

33. The apparatus of claim 21, wherein said music model is one that has been trained to suppress a set of frequencies in said music, said set of frequencies comprising all frequencies below a first frequency, said first frequency having been selected to cause said measured feature-set to be driven towards said target feature- set.

34. The apparatus of claim 21, wherein said music model is one that has been trained to suppress a set of frequencies in said music, said set of frequencies being a union of a first set, which comprises all frequencies below a first frequency, and a second set, which comprises all frequencies above a second frequency, said second frequency being greater than said first frequency, said first and second frequencies having been selected to cause said measured feature-set to be driven towards said target feature-set.

35. The apparatus of claim 21, wherein said feature-extraction circuit receives information from a real-time electroencephalogram.Attorney Docket No. 30263-008W0136. The apparatus of claim 21, wherein said feature-extraction circuit receives information representative of real-time heart rate, heart-rate variability, respiration rate, body temperature, peripheral body temperature, galvanic skin response, changes in galvanic skin response, changes in muscle tension, respiration rate, and respiration volume.

37. The apparatus of claim 21, further comprising a robot, wherein said biofeedback circuitry is within said robot.

38. The apparatus of claim 21, wherein said distributed feedback-circuitry comprises local circuitry that provides said biofeedback circuitry with information concerning real time physiological signals and wherein said biofeedback circuitry is accessible to said local circuitry via a network.

39. A method comprising causing a measured feature-set to be driven towards a target feature-set, wherein causing said measured feature-set to be driven towards said target feature-set comprises receiving, from a subject, real-time physiological information and information that is indicative of a target state- of-consciousness, based on features in said real-time physiological information, forming a conditioning stimulus to cause a state-of-consciousness of said subject, as determined based on said real-time physiological information, to move towards said target state-of-consciousness, and transmitting said conditioning stimulus to said subject, wherein forming said conditioning stimulus comprises using an artificial-intelligence engine comprising a music model to generate music having features that are selected to drive said measured feature-set towards said target feature-set.- SO-

Images