Method and system for extraction and detection of physiological features
First Claim
1. A method of extracting features from multichannel test data, said test data obtained from a physiological sensor suite, said method comprising the steps of:
- (a) extracting feature-absent truth data containing physiological information from said sensor suite;
(b) pre-processing said feature-absent truth data;
(c) further processing said pre-processed data to obtain the parameters of a multichannel model to represent a feature-absent condition associated with said pre-processed data;
(d) forming a feature-absent whitening filter using said parameters of said multichannel model;
(e) extracting said test data containing physiological information from said sensor suite;
(f) pre-processing said test data; and
(g) processing said pre-processed test data with said feature-absent whitening filter to produce an output residual largely containing only white noise and a feature of interest.
1 Assignment
0 Petitions
Accused Products
Abstract
The invention is directed to a method and associated system for the detection of various physiological features in multichannel sensor data. Truth data organized in the form of a vector sequence is used to generate a multichannel state variable model for data that is known to lack physiological features of interest. Such state variable models are obtained using one or more of a host of system identification algorithms. Upon identification of the model parameters, a multichannel inverse filter is derived from the model parameters. The inverse filter can be used to whiten or decorrelate components of sensor data sequences that are not associated with the physiological features of interest. The unwhitened output of the inverse filter then represents the physiological features of interest that can be extracted from the concomitant white noise. On this basis, a number of such filters can be produced for a variety of undesired physiological feature components of the data that occur separately or in combination. The invention is useful for the extraction of known and unknown features of interest from multichannel physiological sensor data.
69 Citations
10 Claims
-
1. A method of extracting features from multichannel test data, said test data obtained from a physiological sensor suite, said method comprising the steps of:
-
(a) extracting feature-absent truth data containing physiological information from said sensor suite; (b) pre-processing said feature-absent truth data; (c) further processing said pre-processed data to obtain the parameters of a multichannel model to represent a feature-absent condition associated with said pre-processed data; (d) forming a feature-absent whitening filter using said parameters of said multichannel model; (e) extracting said test data containing physiological information from said sensor suite; (f) pre-processing said test data; and (g) processing said pre-processed test data with said feature-absent whitening filter to produce an output residual largely containing only white noise and a feature of interest. - View Dependent Claims (2, 3, 4, 5)
-
-
6. A system for extracting features in multichannel test data from a physiological sensor suite, said system comprising:
-
(a) means for extracting feature-absent truth data containing physiological information from said sensor suite; (b) means for pre-processing said feature-absent truth data; (c) means for further processing said pre-processed data to obtain the parameters of a multichannel model to represent a feature-absent condition associated with said pre-processed data, (d) means for forming a feature-absent whitening filter using said parameters of multichannel model; (e) means for extracting said multichannel test data containing physiological information from said sensor suite; (f) means for pre-processing said multichannel test data; and (g) means for processing said pre-processed multichannel test data with said feature-absent whitening filter to produce an output residual largely containing only white noise and a feature of interest. - View Dependent Claims (7, 8, 9, 10)
-
Specification