Surveillance Applications in non-Stationary Acoustic Environments - SALSA
This group has experience in the analysis of sound scenes, both in closed and open spaces, having developed algorithms to integrate a set of techniques for locating, separating and classifying acoustic signals.
During SSPressing project, two different acoustic nodes were designed: a node with no power limitations and high computing features, based on Raspberry Pi; and a low consumption, low cost, self-contained node that does not require external power, with reduced computing capacity, using batteries that are charged by solar cells, and using ultra low power DSPs technology . This solar powered node has a low consumption ARM STM32F4 processor, 4 MEMS microphones, and sub 1-GHz communication capabilities, and it was used to successfully implement a violence detection system.
This group shows deep experience in signal processing algorithms for audiological applications and design of digital hearing aids: algorithms for acoustic environment classification, speech intelligibility enhancement using source separation techniques [Ayl13], optimization of low power consumption algorithms, have been proposed and reported by publications. This team has also shown experience in detecting impulsive sources, such as gunshots [San17b], using array processing algorithms.
-
Acoustical analysis of the physical scenario. The objective is to develop physical models and signal models of the sound produced by target sources (UAVs, gunshots, electric sparks, flaws in materials, etc.), and models of the propagation of sound in the surveillance scenario. The considered frequency bands extend from very low frequency bands where UAVs emit sound, to ultrasound bands, that could be useful for detecting impulsive noise sources (gunshots, electric sparks, etc.), of interest in surveillance application. Propagation waves models will be used later on for self-determination of nodes position in the network using acoustic signals. Sound production models will allow the characterization in time and frequency domains of acoustic signatures of the sound sources.
-
Development of a test-bed for testing surveillance algorithms with DASN. The proposed algorithms will be tested in a test-bed, for which hardware and software improvements will be implemented. New sensors and actuators must be included in the already developed hardware, giving rise to a new version of acoustic node with an extended operating frequency range.
-
Active localization of nodes using non-perceived acoustic signals. Ultrasounds provide a reliable way of self-locating the nodes with high precision, and thus they can be used to help in the array calibration, either using them alone, or in conjunction with other technologies (RSSI, etc.). New promising strategies, such as those based on SLAM algorithm will be explored as well.
-
Application of dynamic acoustic networks for the detection and localization specific events: small UAVs and gunshots. The application of the algorithms to the detection and location of UAVs and gunshots in outdoor environments will be addressed. For indoor environments, the detection of impulsive sounds at homes will be studied, which usually are related to security threats.
- Rosa Zurera, Manuel
Coordinador y Catedrático de Universidad
Read more
- Gil Pita, Roberto
Profesor Titular de Universidad
Read more
- Utrilla Manso, Manuel
- Garín Ciriza, Joaquín
- Jiméz Martínez, Roberto
Associate Professor
Read more
- Llerena Aguilar, Cosme
Read more
- Macías López, Elsa
- Sánchez Hevia, Hector Adrián
- Suarez Sarmiento, Alvaro
Catedrático de Universidad
Read more