Module 5B

Filtering and Smoothing in Data Analysis

Introduction to the Lecture

• The lecture covers the concepts of filtering and smoothing in data reporting.
• The focus is on the differences and connections between smoothing and filtering.
• An introduction to sampling and its effects on data integrity is given.

Sampling and Frequency

• Sampling refers to the process of collecting data points from a continuous signal.
• The sampling rate determines how frequently samples are taken.
  • Examples: Sampling at 100 Hertz (100 samples/second) versus 10 Hertz (10 samples/second).
• Gaps in Sampling: Sampling too infrequently results in missing data, which could affect analysis:
  • Even with high-frequency sampling, gaps may remain.
  • Sufficient sampling frequency is crucial for accurate data collection.

Interpolation

• Definition: Interpolation is a method for estimating unknown values within the range of a discrete set of known data points.
• Importance of Interpolation:
  • It assists in making data appear smoother and more continuous despite actual sampling gaps.
• The accuracy of interpolation improves with higher sampling rates:
  • Example: If multiple points on a line are sampled, a linear regression can predict undetermined data points accurately given a linear relationship.
  • Non-linear relationships might require more complex fitting functions (e.g., polynomial functions).
• Methods of Interpolation: Various techniques exist to interpolate data effectively (specific techniques not detailed in the transcript).
• Example scenario: Interpolation applied when increasing samples from 10 Hertz to 100 Hertz enhances curve approximation.
• Aliasing Error: If the actual frequency of the signal exceeds half the sampling rate, it can distort the interpolated data.

Extrapolation

• Definition: Extrapolation involves estimating values beyond the range of known data points based on an assumed function form.
• Typically used to predict outcomes based on current trends:
  • Example: predicting injury likelihood based on workload and past data patterns.
• Challenges with Extrapolation:
  • Risks arise when making assumptions; the relationship might change beyond known data points, leading to inaccurate predictions.
• The reliability of extrapolation diminishes far from the known data points.

Relation Between Interpolation, Extrapolation, and Smoothing

• Both interpolation and extrapolation contribute to the understanding and processing of data in relation to noise reduction or clarity in representation.
• High sampling rates may result in raw data usable without smoothing; however, filtering or smoothing could be needed when combining different data sources or dealing with inherent noise.

Filtering vs. Smoothing

• Smoothing: Simplifying data and removing erratic variations.
• Filtering: Removing noise from data, often by targeting specific frequencies to eliminate irrelevant signals.
• Interchangeability: While both terms are sometimes used interchangeably, understanding the specific application is crucial (the methodology applied must be precisely documented).

Types of Filtering and Smoothing Techniques

• Three broad approaches to reduce noise or smooth out error in sampled data:
  1. Digital Filters:
     • Most commonly known is the Butterworth filter.
     • Digital filters are designed to manage frequency domains to improve data accuracy.
  2. Mathematical Functions: Utilizing functions such as polynomials to achieve data smoothing.
  3. Frequency Domain Techniques: e.g., Fourier Series Transform:
     • Used both for filtering and understanding frequency peaks within data that may indicate noise.

Importance of Smoothing and Filtering in Dynamics

• Essential for calculating derivatives (e.g., velocity, acceleration) from raw data:
  • High levels of noise in raw measurements can significantly affect the accuracy of derivatives.
  • Understanding how to clean data helps ensure reliability in performance measures, especially in athletic contexts.

Direct and Inverse Dynamics

• Direct Dynamics:
  • Measures both kinetics (forces) and kinematics (motion) directly and combines them to form new variables.
  • Example: Forward dynamics uses measured forces to derive motion metrics.
• Inverse Dynamics:
  • Infers forces or torques based on observed acceleration and known mass.
  • Example: Calculating forces from kinematic data.
• Kinematic measures (displacement, time) are often reported as velocity:
  • Utilizing the finite difference method for calculation.

Impact of Noise on Derivative Calculations

• Noise in original signal data can propagate and amplify through various derivative calculations (e.g., displacement to acceleration).
• The term "garbage in, garbage out" emphasizes the need for clean initial data to prevent erroneous outputs in further calculations.
  • Example visual comparison of raw versus smoothed signals highlighting error magnitude in derivatives.

Impulse and Dynamics Examples

• Impulse is calculated by the product of force and time, represented as the area under a force-time curve.
• From impulse data, velocity and displacement can be derived through integration:
  • High-quality, noise-free initial data is especially important for these integrations to avoid magnifying errors.
• Wording on Filtering and Impulse: Emphasizes the necessity for filtering when employing integration processes that could amplify original signal errors.

Conclusion and Next Steps

• The lecture concludes with a recap of the interplay between data smoothing, filtering, direct/inverse dynamics, and emphasizes the critical nature of collecting accurate data for reliable performance metrics.