Notch Filters without Transient Effects: A Constrained Optimization Design

TL;DR

This paper introduces a constrained least squares optimization method to design notch filters that eliminate transient effects, improving short-duration signal processing such as in wearable biosignal devices.

Contribution

It presents a novel filter design approach that achieves zero transient effects and discusses its limitations and alternatives for nonstationary noise removal.

Findings

Zero-transient notch filters effectively cancel powerline noise.

The proposed filter is equivalent to the Wiener smoother in stationary conditions.

Kalman filter-based methods are suggested for nonstationary noise scenarios.

Abstract

Transient responses are an inherent property of recursive filters due to unknown or incorrectly selected initial conditions. Well-designed stable filters are less affected by transient responses, as the impact of initial conditions diminishes over time. However, applications that require very short data acquisition periods (for example, as short as ten seconds), such as biosignals recorded and processed by wearable technologies, can be significantly impacted by transient effects. But how feasible is it to design filters without transient responses? We propose a well-known filter design scheme based on constrained least squares (CLS) optimization to create zero-transient effect notch filters for powerline noise cancellation. We demonstrate that this filter is equivalent to the optimal Wiener smoother in the stationary case. We also discuss its limitations in removing powerline noise with nonstationary amplitude, where a Kalman filter-based formulation can be used instead.