Introduction to Data Normalization

 

Outline

 

A. Relationship between signals and images

1. 1-D signals - one independent and one dependent variable. Example is a bioelectric signal (electrogram) from a single recording site on the surface of the heart.

2. 2-D signals (images) - two independent and one dependent variable - such as space-x and space-y. Example is a bioelectric signal (electrogram) from multiple recording sites on the surface of the heart.

3. 3-D signals (images) - three independent and one dependent variable - such as space-x axis, space-y axis, and time. Example is a bioelectric signal (electrogram) from multiple recording sites on the surface of the heart with time as the third independent variable.

4. Q-D signals (images) - Q independent and one dependent variable - such as space-x axis, space-y axis, space-z axis and time.

5. Structures with one independent and multiple dependent variables - a set of signals such as blood pressure, heart rate, and temperature which are dependent on time.

B. Normalization

1. amplification - a scale along the y-axis.

2. baseline shift - a translation along the y-axis.

3. stretch or contraction - a scale along the x-axis.

4. phase shift - a translation along the x-axis.

C. Eigenspace

1. Eigenvectors - axes of the ellipsoid. Depends on cluster of N-dimensional points (observations).

2. Eigenvalues - variance along each axis. Depends on cluster of N-dimensional points (observations).

3. Karhunen-Loeve Coefficients - the value of a particular signal along each axis in eigenspace. It is determined as the dor product of signal with eigenvector.