Layered models for context awareness
First Claim
1. A learning system for signal processing comprising a computer processor and memory for executing the following components, the system is recorded on a computer-readable medium and capable of execution by a computer, comprising the:
- N classification layers, each of the N classification layers associated with MN probabilistic models per layer, N and M being integers, respectively;
at least one parameter defined per a respective classification layer, the parameters trained independently at different classification layers;
at least one input stream that is analyzed by the N classification layers, at least one of the input streams sampled according to varying levels of temporal granularity at the respective classification layers;
a plurality of time inputs that are applied as sample inputs to the respective layers, the plurality of time inputs are arranged in a descending order of temporal granularity, the descending order refers to lower levels of the N classification layers being sampled at finer time granularities than higher levels of the N classification layers;
wherein the N classification layers analyze multiple input streams to determine at least one state associated with a human characteristic, the multiple input streams include at least one of audio data, video data, computer activity data, and other contextual activity data, and wherein at least one of the determined states is stored to enable a computer event; and
at least one computer event that is enabled from at least one of the state, and at least one of the computer events is at least one of anticipated, altered, tuned, and adjusted in accordance with a determined state.
2 Assignments
0 Petitions
Accused Products
Abstract
The present invention relates to a system and methodology providing layered probabilistic representations for sensing, learning, and inference from multiple sensory streams at multiple levels of temporal granularity and abstraction. The methods facilitate robustness to subtle changes in environment and enable model adaptation with minimal retraining. An architecture of Layered Hidden Markov Models (LHMMs) can be employed having parameters learned from stream data and at different periods of time, wherein inferences can be determined relating to context and activity from perceptual signals.
161 Citations
42 Claims
-
1. A learning system for signal processing comprising a computer processor and memory for executing the following components, the system is recorded on a computer-readable medium and capable of execution by a computer, comprising the:
-
N classification layers, each of the N classification layers associated with MN probabilistic models per layer, N and M being integers, respectively; at least one parameter defined per a respective classification layer, the parameters trained independently at different classification layers; at least one input stream that is analyzed by the N classification layers, at least one of the input streams sampled according to varying levels of temporal granularity at the respective classification layers; a plurality of time inputs that are applied as sample inputs to the respective layers, the plurality of time inputs are arranged in a descending order of temporal granularity, the descending order refers to lower levels of the N classification layers being sampled at finer time granularities than higher levels of the N classification layers; wherein the N classification layers analyze multiple input streams to determine at least one state associated with a human characteristic, the multiple input streams include at least one of audio data, video data, computer activity data, and other contextual activity data, and wherein at least one of the determined states is stored to enable a computer event; and at least one computer event that is enabled from at least one of the state, and at least one of the computer events is at least one of anticipated, altered, tuned, and adjusted in accordance with a determined state. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)
-
-
17. A computer-implemented method to facilitate learning in a sampled data system, comprising the following computer-executed acts:
-
determining a plurality of parameters for a layered classification system, each layer of the layered classification system includes M associated probabilistic models, M being an integer greater than 1; connecting at least two layers of the layered classification system, via an inferential result from a lower classification layer to a higher classification layer; training the plurality of parameters that are configured on different layers of the layered classification system independently; applying a plurality of time inputs as sample inputs to the layers, the plurality of time inputs are arranged in a descending order of temporal granularity, the descending order refers to lower levels of the N classification layers being sampled at finer time granularities than higher levels of the N classification layers; determining at least one model state from the layered classification system; storing at least one of the determined model states for adjusting a computer event; and adjusting at least one computer event based upon at least one of the determined model states, the computer event comprises at least one of adjusting computer applications, adjusting timing, providing feedback, and changing at least one control. - View Dependent Claims (18, 19, 20, 21, 22, 23)
-
-
24. A computer-implemented method to facilitate signal processing in an intelligent system, comprising the following computer-executed acts:
-
applying a plurality of sensor signals at a base layer of a probabilistic hierarchy; determining at least one feature vector for an intermediary layer of the probabilistic hierarchy; employing a tertiary layer of the probabilistic hierarchy to discriminate data; applying a plurality of time inputs as sample inputs to the base, intermediary and tertiary layers, the plurality of time inputs are arranged in a descending order of temporal granularity, the descending order refers to lower levels of the N classification layers being sampled at finer time granularities than higher levels of the N classification layers; associating longer temporal concepts with subsequent layers of the probabilistic hierarchy, the base layer, the intermediary layer, the tertiary layer and subsequent layers of the probabilistic hierarchy comprise a plurality of probabilistic models; determining at least one model state from the base, intermediary and tertiary layers; storing at least one of the determined model states for adjusting a computer event; and adjusting at least one computer event based upon at least one of the determined model state, the computer event comprises at least one of adjusting computer applications, adjusting timing, providing feedback and changing at least one control. - View Dependent Claims (25, 26, 27, 28, 29, 30)
-
-
31. A computer-implemented method to facilitate real time signal processing, comprising the following computer-executed acts:
-
defining a multi-layer classification system; defining one or more probabilistic models per a respective layer of the multi-layer classification system; defining different sample rates per the respective layer; applying a plurality of time inputs as sample inputs to the respective layers, the plurality of time inputs are arranged in a descending order of temporal granularity, the descending order refers to lower levels of the N classification layers being sampled at finer time granularities than higher levels of the N classification layers; feeding inferential outputs from previous layers to subsequent layers of the multi-layer classification system; inferring multiple states from analytical results derived from previous layers; storing the inferred model states for adjusting a computer event; and adjusting at least one computer event based upon the inferred multiple states, the computer event comprises at least one of adjusting computer applications, adjusting timing, providing feedback, and changing at least one control.
-
-
32. An analysis system comprising a computer processor and memory for executing the following software components, the system is recorded on a computer-readable medium and capable of execution by a computer, comprising:
-
at least one processing component at a lowest layer of a classification system to analyze multiple signal input streams; at least one feature vector indicating results from a portion of the lowest layer; and at least one behavioral model executable on a highest layer of the classification system to infer human behavior based at least in part on the feature vector, the lowest layer, the higher layer and intermediary layers of the classification system comprise a plurality of probabilistic models a plurality of time inputs that are applied as sample inputs to the lowest and highest layers, the plurality of time inputs are arranged in a descending order of temporal granularity, the descending order refers to lower levels of the N classification layers being sampled at finer time granularities than higher levels of the N classification layers, and wherein the inferred human behavior is stared to enable a computer event; and at least one computer event that is enabled from the inferred human behavior, at least one of the computer events is at least one of anticipated, altered, tuned, and adjusted in accordance with the inferred human behavior. - View Dependent Claims (33, 34, 35, 36, 37, 38, 39, 40)
-
-
41. A system to facilitate computer signal processing comprising a computer processor and memory for executing the following software components, the system is recorded on a computer-readable medium and capable of execution by a computer, comprising:
-
means for configuring a multi-layer classification system; means for training one or more probabilistic models per a respective layer of the multi-layer classification system; means for providing various time sample rates per the respective layer; means for applying a plurality of time inputs as sample inputs to the respective layers, the plurality of time inputs are arranged in a descending order of temporal granularity, the descending order refers to lower levels of the N classification layers being sampled at finer time granularities than higher levels of the N classification layers; means for feeding inferential outputs from previous layers to subsequent layers of the multi-layer classification system; means for inferring multiple human states from analytical results derived from previous layers; means for storing the inferred multiple human states for adjusting a computer event; and means for adjusting at least one computer event based upon the inferred multiple human states, the computer event comprises at least one of adjusting computer applications, adjusting timing, providing feedback, and changing at least one control.
-
-
42. A signal to communicate sensor data between at least two nodes system for signal processing, comprising a computer processor and memory for executing the following components, the signal is recorded on a computer-readable medium and capable of execution by a computer, comprising:
a first data packet comprising; an N layered classification component employed to communicate probabilistic results between at least two layers of N layered classification system; and at least one parameter defined per a respective classification layer, the parameters trained independently at different classification layers; and at least one input stream that is analyzed by the N classification layers, at least one of the input streams is sampled according to varying levels of temporal granularity at the respective classification layers, the N layered classification component and the different classification layers comprise M probabilistic models, M and N being integers greater than 1; a plurality of time inputs that are applied as sample inputs to the respective layers, the plurality of time inputs are arranged in a descending order of temporal granularity, the descending order refers to lower levels of the N classification layers being sampled at finer time granularities than higher levels of the N classification layers; wherein the N classification layers analyze multiple input streams to determine at least one state associated with a human characteristic, the multiple input streams include at least one of audio data, video data, computer activity data, and other contextual activity data, and wherein at least one of the states is stored to enable a computer event; and at least one computer event that is enabled from at least one of the states, and at least one of the computer events is at least one of anticipated, altered, tuned, and adjusted in accordance with a determined state.
Specification