value()

double MayaFlux::Kinesis::Stochastic::Estimate::value ( ) const

inline

Current filtered value, the cleaned counterpart to the raw sample.

This is the value a caller should differentiate, not the raw sample last passed to update(). Estimate exists to turn a noisy arriving stream into something Kinesis::Differential can safely take derivatives of: differentiation amplifies noise, so feeding Differential's HistoryBuffer-based functions a raw high-resolution device stream directly produces a velocity/acceleration/jerk signal dominated by sensor noise rather than motion.

What "filtered" means depends on the active model:

• ROLLING_VARIANCE / MEDIAN_ABSOLUTE_DEVIATION: the window mean or median, a straightforward smoothing filter
• EWM_VARIANCE: the exponentially weighted running mean
• QUIET_PERIOD_FLOOR: the last quiet-period mean while the stream looks calm, but the raw sample itself while the stream looks active, since lagging behind a stale mean during real motion would corrupt exactly the transient a caller most wants intact
• TREND: the trend line's value at the newest sample, tracking directional motion while smoothing residual jitter around it

Definition at line 258 of file Estimate.hpp.

{ return m_state.filtered_value; }