ESPRIT (Estimation of Signal Parameters via Rotational Invariance Techniques) is a subspace-based spectral estimation algorithm that exploits the rotational invariance property between two displaced but identical subarrays of an antenna array to efficiently estimate the directions of arrival (DoA) or other signal parameters such as time delay and frequency without requiring an exhaustive search over a parameter space, unlike MUSIC. In WiFi/CSI sensing, ESPRIT is important because it enables high-resolution, computationally efficient estimation of multipath component parameters — including angle, range, and velocity — directly from channel state information, making it well suited for real-time localization and sensing applications. Key variants include TLS-ESPRIT (Total Least Squares ESPRIT), which improves robustness to noise, and unitary ESPRIT, which reduces computational cost by transforming the problem into real-valued arithmetic, as well as extensions to 2D-ESPRIT for joint azimuth and elevation estimation relevant to multi-antenna 802.11bf sensing systems.
Dictionary term