Audio, speech and image processing are at the core of the Signal Processing research line.
The audio processing research line is focused on audio event detection in three different frameworks. The first one is SmartCities, where the investigations are focused on detecting outdoor anomalous noise events in the noise level measurements of road traffic, in order to avoid the impact of these non-traffic related events to the tailoring of the noise maps.
The second one is carried out at home within the Ambient Assisted living paradigm. It is oriented to detect indoor acoustic events derived from everyday activities – water, glass breaking, dog barking - for behavior and surveillance remote monitoring.
The third one is climate change through the acoustic detection of birdsongs of endangered species. In all them, the main challenge is to improve machine learning and feature extraction to improve the performance of the envisioned systems. As a general goal, we are also working towards real-time processing implementations running over a low-cost hardware platforms such as FPGAs and GPUs.
The speech processing area is focused on analyzing and synthesizing expressive speech for storytelling. We take into account the characterization and transformation of this characteristic speaking style to convey the appropriate expressiveness to Text-to-Speech (TTS) synthesis systems, even adding the capability of producing a character’s voice singing. The speech processing area is also involved in including expressivity in the numerical voice generation, together with the Acoustics research line of GTM.
Finally, in terms of image processing, research work will be oriented to improve the specialization on R&D at the frontier between engineering and biology fields. Two main areas of research will be tackled: the development of accurate 2D and 3D face modelling for the diagnosis and prognosis of mental diseases with genetic causes, and the development of automatic tools for the measurement of biologically relevant plant traits at both microscopic and macroscopic level.
Team
