Apparatus and method for classifying orientation of a body of a mammal

1. An apparatus for providing a classification of body orientation of a mammal in static and/or dynamic states, said apparatus including:

two position sensors located on an upper body of the mammal to measure body position relative to a frame of reference and to provide first data indicative of said body position;

a memory device coupled to the two position sensors and arranged for storing said first data at least temporarily;

a processor to receive said first data and to process said first data to provide said classification of body orientation in said static and/or dynamic states, wherein said processor implements a Body Orientation Classification (BOC) algorithm for evaluating said body orientation, said BOC algorithm including:

a dynamic classifier configured to identify dynamic movement based on identification of a cyclic pattern in the first data received from the two position sensors;

a static classifier configured to identify posture based on the first data received from the two position sensors mapping to predetermined values; and

a transition classifier configured to identify a pattern of transition in response to a changing state as identified by the static classifier and based on the first data received from the two position sensors comprising a moving window of position values in real time correlating to predefined statistics from observed data, wherein the correlation serves as a confirmation or refutation of the changing state; and

wherein said BOC algorithm updates the predefined statistics based on said first data for subsequent classification of subsequent body position data in a feedback loop, and further outputs a status of the changing state.

2. The apparatus according to claim 1 , wherein said two position sensors are located on a spine of the mammal.

3. The apparatus according to claim 1 , wherein said processing is performed in real time to enhance accuracy of said classification.

4. The apparatus according to claim 1 , wherein said algorithm is adapted to evaluate said body orientation based on assigning class signatures to primary body orientations including standing, sitting and lying down, and wherein said algorithm includes a logic tree to identify posture based on recognition of a signature or pattern relating to each posture.

5. The apparatus according to claim 4 , wherein the signature or pattern relating to each posture is determined by a set of rules acquired during a training phase, and wherein the signature or pattern relating to each posture is modified by means of an unsupervised classifier.

6. The apparatus according to claim 5 , wherein said unsupervised classifier clusters said first data and updates the signature or pattern based on distance to a center of a corresponding cluster in a d-dimensional space wherein d is the number of variables that defines the signature.

7. The apparatus according to claim 1 , wherein said two position sensors include at least one of an accelerometer, a gyroscope and a magnetometer, and wherein said two position sensors are adapted to measure angular displacement along three orthogonal axes.

8. The apparatus according to claim 1 , wherein said first data is used to derive one or more of:

displacement in an extension flexion plane;

displacement in a lateral flexion plane; and

body rotation of said mammal.

9. The apparatus according to claim 1 , wherein said two position sensors include at least one A to D converter for converting analog data to a digital domain, and wherein said A to D conversion takes place prior to storing said first data.

10. A method for providing a classification of body orientation of a mammal in static and/or dynamic states, said method including:

measuring body position of the mammal relative to a frame of reference, said measuring being performed by two position sensors located on an upper body of the mammal to provide said classification of body orientation in said static and/or dynamic states;

providing first data indicative of said body position from said two position sensors;

storing said first data at least temporarily in a memory device;

processing said first data by a processor to provide said classification of body orientation in said static and/or dynamic states by executing a Body Orientation Classification (BOC) algorithm for evaluating said body orientation, said BOC algorithm including:

a dynamic classifier configured to identify dynamic movement based on identification of a cyclic pattern in the first data received from the two position sensors;

a static classifier configured to identify posture based on the first data received from the two position sensors mapping to predetermined values; and

a transition classifier configured to identify a pattern of transition in response to a changing state as identified by the static classifier and based on the first data received from the two position sensors comprising a moving window of position values in real time correlating to predefined statistics from observed data, wherein the correlation serves as a confirmation or refutation of the changing state;

updating, by the processor, the predefined statistics based on said first data for subsequent classification of subsequent body position data in a feedback loop; and

outputting, by the processor, a status of the changing state.

11. The method according to claim 10 , wherein said two position sensors are located on a spine of a mammal.

12. The method according to claim 10 , wherein said processing is performed in real time to enhance accuracy of said classification.

13. The method according to claim 10 , wherein said algorithm is adapted to evaluate said body orientation based on assigning class signatures to primary body orientations including standing, sitting and lying down, and wherein said algorithm includes a logic tree to identify posture based on recognition of a signature or pattern relating to each posture.

14. The method according to claim 13 , wherein the signature or pattern relating to each posture is determined by a set of rules acquired during a training phase, and including modifying the signature or pattern relating to each posture by means of an unsupervised classifier.

15. The method according to claim 14 , wherein said unsupervised classifier clusters said first data and updates the signature or pattern based on distance to a center of a corresponding cluster in a d-dimensional space wherein d is the number of variables that defines the signature.

16. The method according to claim 10 , wherein said two position sensors include at least one of an accelerometer, a gyroscope and a magnetometer, and wherein said two position sensors are adapted to measure angular displacement along three orthogonal axes.

17. The method according to claim 10 , wherein said first data is used to derive one or more of:

displacement in a lateral flexion plane;

displacement in an extension flexion plane; and

body rotation of said mammal.

18. The method according to claim 10 , wherein each step of measuring includes converting analog data to a digital domain, and wherein the converting of data to the digital domain takes place prior to storing said first data.