Automatic speech recognition and natural language processing
We have research and applied the last techniques and deep neural networks architectures for automatic speech recognition to develop models for different domains in Spanish, English, Basque, German, Portuguese and French among others. For doing this we have extensively explored E2E architectures including DeepSpeech2, Jasper, Quartznet and semisupervised learning methods including wav2vec2.0. They have been applied to challenging domains including low resource edge devices, to Federated Learning systems in H2020 European Projects (GRACE) and to critical environments suchs a the EU project APPRAISE .
Computational linguistics, linguistic laws and complexity in communication
I am interested in the study of statistical patterns (including linguistic laws) in language and general communication systems including: self organized criticality in voice, linguistic laws using physical units and comparative studies with animals. During last years we have published the first extensive study of linguistic laws using speech magnitudes and approached the research on complexity of communication through an interdisciplinary collaboration including applied physics, linguistics, cognitive sciences and applied mathematics.
Application of artificial intelligence and applied physics to medical and clinical research
I am interested in applying artificial intelligence algorithms and fundamental physics to the medical and clinical applications. Particularly we are exploring early detection of Alzheimer's disease through non-invasive low cost applications based on speech characteristics and by using feature extraction with deep neural networks. Additionally, we have published new techniques on automatic speech recognition based on semisupervised learning methods that could be used to develop personalized apps in the rehabilitation of patients with aphasia.
Data science, time serie forecasting and deep neural networks in industry 4.0
Among my academic work I have collaborated with Vicomtech and Agrowingdata working with different partners and developing technological solutions based on the last advances of artificial intelligence and applied physics. Specifically, I have carried out technological and knowledge transfer from the academic sector to industry managing technological working packages in multipartner european and national projects.
Multifractal complexity characterization of 2D and 3D soil pore images
I have worked on characterising the complex structure of pore distribution in 2D and 3D soil images applying multifractal analysis, lacunarity or configuration entropy analysis. We have shown that the soil pore structure multifractal spectrums reflect the complexity better than the analysis of binarized soil CT scan images. Additionally, we showed that it is necessary to study the multifractal characterization in 3D as not to be influenced by gravitational effects
