Blind identification of short time sources in multi path and multi physics environment
Abstract
The basic-research focus of this project is the development of new mixing models and algorithms to solve the problem of blind extraction of an independent source from a mixture of signals observed by a set of sensors. The central motivation is to parameterize themodels in order to ensure more accurate and robust system identification than state-of-the-art models. This will be achieved by investigating reduced parameterizations of the mixing model with respect to the structure of the sensor systems and the properties of the environment through which the vibrations propagate. Time-variant systems and situations when the amount of observed data is fundamentally limited will be of particular interest.Blind system identification (BSI) has been applied successfully in fields ranging from cosmology to bio-medicine to room acoustics. This abstract mathematical solution ensures a wide impact on many fields. The practical focus in the project is in identifying changes in mechanical systems consisting of noise/vibration generating parts. These changes can be caused by a malfunction of one component, which can lead to the failure of the whole system. Early identification and elimination of such faults can lead to resource savings. The models and algorithms developed in this project have the potential to advance the use of BSI in this application area. While the use of BSI has been well researched in acoustics and audio signal processing,structural vibrations have many other specific properties that can be anticipated and included in the model. The team will collaborate closely with US Navy researchers at the Naval Surface Warfare Center # Carderock Division to exchange insights, algorithms, and sample problem sets. By such collaboration, the algorithms developed under this project can be directly transitioned to the US Navy.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 11, 2023
- Source ID
- N629092312084
Entities
People
- Jaroslav Čmejla
Organizations
- Office of Naval Research
- Technical University of Liberec
- United States Navy