# US 2014/0207401 A1 · Apparatus for recognizing motion feature of user, method for generating orthogonal non-negative matrix factorization (onmf)-based basis matrix, and method for generating orthogonal semi-supervised non-negative matrix factorization (ossnmf)-based basis matrix

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2014/0207401 A1
- Document ID: US20140207401A1
- Status: Published application
- Kind: A1
- Application number: 14/159,790
- Filed: 2014-01-21
- Published: 2014-07-24
- Earliest claimed: 2013-01-18
- Assignees: Postech Academy-Industry Foundation; Samsung Electronics Co., Ltd.
- Applicants: Postech Academy-Industry Foundation; Samsung Electronics Co., Ltd.
- Inventors: Jae-Mo SUNG; Won-Keun KONG; Sae-Hun KIM; Seung-Jin CHOI
- Web page: https://patentgrep.com/patent/US20140207401A1

## CPC

- G06F 17/16: Matrix or vector computation {, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization (matrix transposition G06F7/78)}
- G01C 21/16: by integrating acceleration or speed, i.e. inertial navigation
- G01C 25/005: {initial alignment, calibration or starting-up of inertial devices}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G01C: MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY (measuring liquid level G01F; radio navigation, determining distance or velocity by use of propagation effects, e.g. Doppler effects, propagation time, of radio waves, analogous arrangements using other waves G01S)

## Abstract

An apparatus for recognizing a user's motions based on sensor information and label information, a method for establishing an ONMF-based basis matrix, and a method for establishing an OSSNMF-based basis matrix are provided, where the basis matrices are used to extract motion features of the user. The apparatus for recognizing the user's motions may include a feature vector extractor configured to multiply a transposed matrix of an orthogonalized basis matrix by a sensor data matrix of frequency domain sensor data acquired from sensors to extract an ONMF-based feature vector and a multi-class classifier configured to use the extracted ONMF-based feature vector to classify the user's motion into a type from among types of a user's motions.

## Claims

### Claim 1

An apparatus for recognizing a user's motions, comprising:
a feature vector extractor configured to multiply a transposed matrix of an orthogonalized basis matrix by a sensor data matrix of frequency domain sensor data acquired from sensors to extract an ONMF (Orthogonal Non-negative Matrix Factorization)-based feature vector; and
a multi-class classifier configured to use the extracted ONMF-based feature vector to classify the user's motion into a type from among types of a user's motions.

### Claim 2 (depends on 1)

The apparatus of claim 1, further comprising:
a sensor data calibrator configured to calibrate at least a portion of the frequency domain sensor data,
wherein the frequency domain sensor data is converted to have a global coordinate system by the sensor data calibrator using additional sensor data.

### Claim 3 (depends on 1)

The apparatus of claim 1, wherein the feature vector extractor is further configured to extract a non-ONMF-based feature vector comprising at least one of frequency domain sensor data and time domain sensor data, and integrate the ONMF-based feature vector and the non-ONMF-based feature vector to generate an integrated feature vector, and
the multi-class classifier is further configured to use the integrated feature vector to classify the user's motion into a type from among types of a user's motions.

### Claim 4 (depends on 1)

The apparatus of claim 1, wherein a basis matrix and an encoding matrix corresponding to the basis matrix are converged to update the orthogonalized basis matrix.

### Claim 5 (depends on 1)

The apparatus of claim 1, wherein the orthogonalized basis matrix is an approximately orthogonal matrix produced using Stiefel manifolds.

### Claim 6

An apparatus for recognizing a user's motions, comprising:
a first feature vector extractor configured to multiply a sensor data matrix comprising frequency domain sensor data acquired from sensors by a transposed matrix of a first basis matrix orthogonalized with respect to the sensor data matrix to extract a first feature vector;
a second feature vector extractor configured to multiply a label matrix comprising label information by a transposed matrix of a second basis matrix of the label matrix to extract a second feature vector;
an integrated feature vector generator configured to integrate the first feature vector and the second feature vector to generate an integrated feature vector; and
a multi-class classifier configured to use the integrated feature vector to classify the user's motion into a type from among types of a user's motions.

### Claim 7 (depends on 6)

The apparatus of claim 6, wherein the second basis matrix has the same encoding matrix as the first basis matrix.

### Claim 8 (depends on 7)

The apparatus of the claim 7, wherein the first basis matrix and an encoding matrix corresponding to the first basis matrix are converged to update the first basis matrix.

### Claim 9

A method for generating an Orthogonal Non-negative Matrix Factorization (ONMF)-based basis matrix for extracting motion features of a user, comprising:
obtaining a sensor data matrix comprising sensor data acquired from sensors;
extracting an integrated feature vector from the sensor data matrix comprising an ONMF-based feature vector; and
generating a basis matrix of the integrated feature vector.

### Claim 10 (depends on 9)

The method of claim 9, wherein the generating of the basis matrix further comprises:
randomly initializing the basis matrix and an encoding matrix of the basis matrix;
updating an initialized basis matrix and an initialized encoding matrix; and
orthogonalizing the basis matrix.

### Claim 11 (depends on 10)

The method of claim 10, wherein the basis matrix is updated using Stiefel manifolds.

### Claim 12

A method for generating an Orthogonal Semi-Supervised Non-negative Matrix Factorization (OSSNMF)-based basis matrix for extracting motion features of a user comprising:
acquiring sensor data measured from a plurality of sensors and a spectrogram of the sensor data;
generating an integrated feature vector comprising an OSSNMF-based feature vector based on the spectrogram and label data;
randomly initializing a first basis matrix of the sensor data, a second basis matrix of the label data, and an encoding matrix;
updating an initialized first basis matrix, an initialized second basis matrix and an initialized encoding matrix; and
generating a basis matrix of the integrated feature vector.

### Claim 13 (depends on 12)

The method of claim 12, further comprising:
terminating the updating of the initialized matrices in response to the first basis matrix converging to a stable value.

### Claim 14 (depends on 12)

The method of claim 12, wherein the updating further comprises L2-normalizing the first basis matrix.

### Claim 15

An apparatus for generating an Orthogonal Non-negative Matrix Factorization (ONMF)-based basis matrix for extracting motion features of a user, comprising:
sensors; and
a feature vector extractor configured to:
obtain sensor data from the sensors;
extract an integrated feature vector from the sensor data comprising an ONMF-based feature vector; and
generate a basis matrix of the integrated feature vector.

### Claim 16 (depends on 15)

The apparatus of claim 15, further comprising a sensor data calibrator configured to calibrate at least a portion of the sensor data.

### Claim 17 (depends on 15)

The apparatus of claim 15, wherein the integrated feature vector additionally comprises a non-ONMF-based feature vector comprising at least one of frequency domain sensor data and time domain sensor data.

### Claim 18 (depends on 15)

The apparatus of claim 15, wherein the basis matrix is an approximately orthogonal matrix produced using Stiefel manifolds.
