Direction of arrival (DOA) estimation is a well-studied problem in signal processing with applications in radar, sonar, acoustics, seismology, and wireless communications. It involves estimating the direction from which a signal arrives at a sensor array, for example to locate a missing person or animal from a wireless tag or the sound they produce. Traditionally, DOA estimation relies on time difference of arrival (TDOA) measurements between spatially separated sensors. However, accurately estimating these time differences can be challenging and often requires sufficiently large sensor arrays.

The main contribution in this thesis is a DOA estimation method based on received signal power rather than TDOA measurements. It exploits the directional sensitivity of microphones and antennas, modeled using a Fourier series (FS) representation that provides a compact, parametric description of sensor responses over the periodic range of DOA angles. This model is used to derive the proposed estimator and analyze its theoretical performance using the Cramér-Rao lower bound (CRLB).

The proposed method is evaluated in the acoustic domain using a range of signals, including a gunshot, an elephant trumpet, a siren, and a female scream. The results show that the directional sensitivity depends strongly on the frequency content of the signal, which can be exploited for accurate DOA estimation. Across the investigated natural signals, the method matches or outperforms established approaches, including multiple signal classification (MUSIC) and minimum variance distortionless response (MVDR), in both anechoic and outdoor environments.

A key advantage is that the method requires no TDOA measurements, enabling smaller and portable sensor devices. Its main limitation is that the directional sensitivity must be determined by calibration in a controlled environment. The calibration is nevertheless shown to be robust across different environmental conditions and over time.

The framework is further extended to measurements subject to detection thresholds. Rather than discarding signals that fall below the threshold, the proposed estimator jointly exploits detections and missed detections through a likelihood function that accounts for both outcomes. The formulation combines the FS sensor model with a known detection threshold and is independent of the sensing modality, making it applicable whenever the sensor response and detection threshold are known. The framework is validated using Bluetooth low energy (BLE) signals from directional antennas, where receiver thresholding provides a natural example of incomplete measurements. The proposed method consistently achieves higher DOA estimation accuracy than a baseline that ignores missed detections, with the advantage increasing at higher missed detection rates.

Finally, the framework is integrated into the Bernoulli filter (BF) formulation for recursive tracking in the presence of missed detections, false alarms, and uncertainty about target existence. Simulations show accurate tracking even at increasing missed detection rates, while a baseline that assumes a constant detection probability degrades. Classical TDOA-based estimation and tracking are additionally illustrated using seismic measurements of elephant footsteps.

Together, these contributions form a general DOA estimation and tracking framework using directionally sensitive sensors, including measurements subject to detection thresholds. The framework is demonstrated using acoustic and wireless sensing and enables small sensor arrays without TDOA measurements while maintaining accurate estimation and tracking even when a large fraction of potential measurements are lost.