A5043481820- EEG and Brain-Computer Interfaces
- Authorship Attribution and Profiling
- Speech and Audio Processing
- Video Analysis and Summarization
- Logic, Reasoning, and Knowledge
- Advanced Memory and Neural Computing
- AI-based Problem Solving and Planning
- Natural Language Processing Techniques
- Advanced Image and Video Retrieval Techniques
- Indoor and Outdoor Localization Technologies
- Topic Modeling
- Neural Networks and Applications
- Advanced Adaptive Filtering Techniques
- Emotion and Mood Recognition
- Modular Robots and Swarm Intelligence
- Recommender Systems and Techniques
- Music and Audio Processing
- Speech Recognition and Synthesis
- Human Pose and Action Recognition
- Spam and Phishing Detection
- Image Retrieval and Classification Techniques
- Neural dynamics and brain function
- Gaze Tracking and Assistive Technology
- Multi-Agent Systems and Negotiation
- Semantic Web and Ontologies
Universidad Nacional Autónoma de México
2016-2025
Universidad Autónoma Monterrey
2021-2022
Institute of Mathematical Sciences
2018
University of Electro-Communications
2009-2011
Most emotion recognition systems still present limited applicability to new users due the inter-subject variability of electroencephalography (EEG) signals. Although domain adaptation methods have been adopted tackle this problem, most methodologies deal with unlabeled data from a target subject. However, few labeled samples subject could also be included boost cross-subject recognition. In paper, we semi-supervised framework align joint distributions subjects, assuming that fine-grained...
Multi-modal classifiers for emotion recognition have become prominent, as the emotional states of subjects can be more comprehensively inferred from Electroencephalogram (EEG) signals and eye movements. However, existing experience a decrease in performance due to distribution shift when applied new users. Unsupervised domain adaptation (UDA) emerges solution address between by learning shared latent feature space. Nevertheless, most UDA approaches focus on single modality, while multi-modal...
Abstract We employ the XGBoost machine learning (ML) method for morphological classification of galaxies into two (early-type, late-type) and five (E, S0–S0a, Sa–Sb, Sbc–Scd, Sd–Irr) classes, using a combination non-parametric (C, A, S, AS, Gini, M20, c5090), parametric (Sérsic index, n), geometric (axial ratio, BA), global colour (g − i, u r, i), gradient (Δ(g i)), asymmetry (ΔA9050) information, all estimated local galaxy sample (z < 0.15) compiled from Sloan Digital Sky Survey...
We present a neural model for generating short stories from image sequences, which extends the description by Vinyals et al. (Vinyals al., 2015). This extension relies on an encoder LSTM to compute context vector of each story sequence. is used as first state multiple independent decoder LSTMs, generates portion corresponding in sequence taking embedding input. Our showed competitive results with METEOR metric and human ratings internal track Visual Storytelling Challenge 2018.
Although the capacity of deep generative models for image generation, such as Diffusion Models (DMs) and Generative Adversarial Networks (GANs), has dramatically improved in recent years, much their success can be attributed to computationally expensive architectures. This limited adoption use research laboratories companies with large resources, while significantly raising carbon footprint training, fine-tuning, inference. In this work, we present LadaGAN, an efficient adversarial network...
In this paper, we present a concept of service robot and framework for its functional specification implementation. The discussion is grounded in Newell's system levels hierarchy which suggests organizing robotics research three different layers, corresponding to Marr's computational, algorithmic implementation levels, as follows: (1) the proper, subject (2) perception action algorithms, (3) systems programming level. articulated practice through introduction conceptual model particular...
In this paper we present SitLog: a declarative situation-oriented logical language for programming situated service robot tasks. The formalism is task and domain independent, can be used in wide variety of settings. SitLog also seen as behaviour engineering specification interpretation to support action selection by autonomous agents during the execution complex combines recursive transition network formalism, extended with functions express dynamic contextualized structures, functional...
Social media platforms, such as Twitter (now X), are a major source of communication. Identifying communicative intentions is useful, it encapsulates the latent motivations that drive text creation. This intention also helpful in understanding message, context, and audience. study proposes method for detecting tweets using Jakobson’s language functions. We constructed meticulously annotated dataset, drawing from extensive RepLab2013 corpus. Our dataset underwent rigorous scrutiny by...
Sound source localization is important in human interaction, such as locating the origin of long-distance calls or facing other humans while a conversation. It interest to apply functionality core human-robot interaction (HRI) and investigate its benefits, if any. In this paper, we propose three strategies for how integrate multiple directions-of-arrival (multi-DOA) estimation with common scenario, which robot acts waiter applying audio localization. The proposed are: a) locates from users...
Abstract Natural memories are associative, declarative and distributed, memory retrieval is a constructive operation. In addition, cues of objects that not contained in the rejected directly. Symbolic computing resemble natural their character, information can be stored recovered explicitly; however, they reproductive rather than constructive, lack associative distributed properties. Sub-symbolic developed within connectionist or artificial neural networks paradigm but property, capability...
Estimating the directions of arrival (DOAs) multiple simultaneous mobile sound sources is an important step for various audio signal processing applications. In this contribution, we present approach that improves upon our previous work now able to estimate DOAs speech sources, while being light in resources, both hardware-wise (only using three microphones) and software-wise. This takes advantage fact do not completely overlap each other. To evaluate performance approach, a multi-DOA...
We present a scalable approach to automatically discovering particular objects (as opposed object categories) from set of images. The basic idea is search for local image features that consistently appear in the same images under assumption such co-occurring underlie object. first represent each as visual words (vector quantized features) and construct an inverted file memorize which word appears. Then, our discovery method proceeds by searching extracting sets whose elements tend images;...
We present the use of direction arrival (DOA) sound sources as an index during interaction between humans and service robots. These indices follow notion defined by theory interpretation signs Peirce. This establishes a strong physical relation (DOAs) objects being signified in specific contexts. With this mind, we have modeled call at distance to robot indexical nature. can be later interpreted position user herself/himself. The emitter is formalized our framework development robots based...
This paper presents a novel approach to discovering particular objects from set of unannotated images. We aim find discriminative feature sets that can effectively represent object classes (as opposed categories). achieve this by mining correlated visual word the bag-of-features model. Specifically, we consider belongs same class if all its words consistently occur together in image. To efficiently such apply Min-LSH occurrence vector each word. An agglomerative hierarchical clustering is...
The Acoustic Interactions for Robot Audition corpus is introduced research on sound source localization and separation, multi-user speech recognition. Its aim to evaluate train techniques, as well Auditory Scene Analysis in general. It was recorded six real-life environments with different noise presence reverberation time, using two array configurations: an equilateral triangle, a three-dimensional 16-microphone set over hollow plastic body. includes clean data static sources tracking...