A5103257592- Speech and Audio Processing
- Speech Recognition and Synthesis
- Music and Audio Processing
- Social Robot Interaction and HRI
- Advanced Data Compression Techniques
- Speech and dialogue systems
- Human Pose and Action Recognition
- Robotic Path Planning Algorithms
- Robot Manipulation and Learning
- Modular Robots and Swarm Intelligence
- Advanced Adaptive Filtering Techniques
- Digital Filter Design and Implementation
- Robotics and Automated Systems
- Robotics and Sensor-Based Localization
- Visual Attention and Saliency Detection
- Technology Use by Older Adults
- Natural Language Processing Techniques
- Hand Gesture Recognition Systems
- Robotic Locomotion and Control
- Blind Source Separation Techniques
- Video Surveillance and Tracking Methods
- Context-Aware Activity Recognition Systems
- Reinforcement Learning in Robotics
- Precipitation Measurement and Analysis
- Advanced Image and Video Retrieval Techniques
Universidad de Málaga
2014-2024
Universitat Politècnica de Catalunya
2007-2024
Hospital Universitario Virgen de la Arrixaca
2000-2012
Universidad de Murcia
2012
Universidad de Granada
1996-2006
Universities UK
1998-1999
University of York
1998
This paper describes a method of compensating for nonlinear distortions in speech representation caused by noise. The described here is based on the histogram equalization often used digital image processing. Histogram applied to each component feature vector order improve robustness recognition systems. how proposed can be robust and it compared with other compensation techniques. experiments, including results AURORA II framework, demonstrate effectiveness when either alone or combination
Currently, there are technology barriers inhibiting speech processing systems that work in extremely noisy conditions from meeting the demands of modern applications. This letter presents a new voice activity detector (VAD) for improving detection robustness environments and performance recognition systems. The algorithm defines an optimum likelihood ratio test (LRT) involving multiple independent observations. so-defined decision rule reports significant improvements speech/nonspeech...
Background. It is not known whether the pig liver capable of functioning efficiently when transplanted into a primate, neither there experience in transplanting from transgenic pigs expressing human complement regulator decay accelerating factor (h-DAF) baboon. The objective this study was to determine porcine would support metabolic functions non-human primates and establish effect hDAF expression prevention hyperacute rejection livers primates. Methods. Five orthotopic xenotransplants...
Since its introduction in 1974 by Ahmed et al., the discrete cosine transform (DCT) has become a significant tool many areas of digital signal processing, especially compression. There exist eight types transforms (DCTs). We obtain DCTs as complete orthonormal set eigenvectors generated general form matrices same way Fourier (DFT) can be obtained an arbitrary circulant matrix. These decomposed sum symmetric Toeplitz matrix plus Hankel or close to scaled some constant factors. also show that...
An effective voice activity detection (VAD) algorithm is proposed for improving speech recognition performance in noisy environments. The approach based on the determination of speech/nonspeech divergence by means specialized order statistics filters (OSFs) working subband log-energies. This differs from many others way decision rule formulated. Instead making current frame, it uses OSFs log-energies which significantly reduces error probability when discriminating nonspeech a signal. Clear...
This letter shows an innovative voice activity detector (VAD) based on the Kullback-Leibler (KL) divergence measure. The algorithm is evaluated in context of recently approved ETSI standard for distributed speech recognition (DSR). VAD uses long-term information noisy signal order to define a more robust decision rule yielding high accuracy. mel-scaled filter bank log-energies (FBE) are modeled by means Gaussian distributions, and symmetric KL used estimation distance between noise...
The noise usually produces a non-linear distortion of the feature space considered for Automatic Speech Recognition. This causes mismatch between training and recognition conditions which significantly degrades performance speech recognizers. In this contribution we analyze effect additive over cepstral based representations compare several approaches to compensate effect. We discuss importance non-linearities introduced by propose method (based on histogram equalization technique)...
Learning by imitation is a natural and intuitive way to teach social robots new behaviors. While these learning systems can use different sensory inputs, vision often their main or even only source of input data. However, while many vision-based robot (RLbI) architectures have been proposed in the last decade, they may be difficult compare due absence common, structured description. The first contribution this survey definition set standard components that used describe any RLbI...
The use of new assistive technologies in general, and Socially Assistive Robots (SAR) particular is becoming increasingly common for supporting people’s health well-being. However, it still faces many issues regarding long-term adherence, acceptability utility. Most these are due to design processes that insufficiently take into account the needs, preferences values intended users. Other related currently very limited amount evaluations, performed real world settings, SAR. This study...
This letter presents a new segmental nonlinear feature normalization algorithm to improve the robustness of speech recognition systems against variations acoustic environment. An experimental study best delay-performance tradeoff is conducted within AURORA-2 framework, and comparison with two commonly used algorithms presented. Computationally efficient based on order statistics are also One them linear interpolation between sampling quantiles, other one point estimation probability...
Motion capture systems have recently experienced a strong evolution. New cheap depth sensors and open source frameworks, such as OpenNI, allow for perceiving human motion on-line without using invasive systems. However, these proposals do not evaluate the validity of obtained poses. This paper addresses this issue model-based pose generator to complement OpenNI tracker. The proposed system enforces kinematics constraints, eliminates odd poses filters sensor noise, while learning real...
In recent years, commercial and research interest in service robots working everyday environments has grown. These devices are expected to move autonomously crowded environments, maximizing not only movement efficiency safety parameters, but also social acceptability. Extending traditional path planning modules with socially aware criteria, while maintaining fast algorithms capable of reacting human behavior without causing discomfort, can be a complex challenge. Solving this challenge...
Over the past decades, number of robots deployed in museums, trade shows and exhibitions have grown steadily. This new application domain has become a key research topic robotics community. Therefore, are designed to interact with people these domains, using natural intuitive channels. Visual perception speech processing be considered for robots, as they should able detect their environment, recognize degree accessibility engage them social conversations. They also need safely navigate...
One of the main issues within field social robotics is to endow robots with ability direct attention people whom they are interacting. Different approaches follow bio-inspired mechanisms, merging audio and visual cues localize a person using multiple sensors. However, most these fusion mechanisms have been used in fixed systems, such as those video-conference rooms, thus, may incur difficulties when constrained sensors which robot can be equipped. Besides, scope interactive autonomous...
The use of new assistive technologies in general, and Socially Assistive Robots (SARs) particular, is becoming increasingly common for supporting people’s health well-being. However, it still faces many issues regarding long-term adherence, acceptability utility. Most these are due to design processes that insufficiently take into account the needs, preferences values intended users. Other related currently very limited amount evaluations, performed real-world settings, SARs. This study...
The aging of the population in developed and developing countries, together with degree maturity reached by certain technologies, means that design care environments for elderly a high technological innovation is now being seriously considered. Assistive daily living (Ambient Assisted Living, AAL) include deployment sensors actuators home or residence where person to be cared lives so that, help necessary computational management decision-making mechanisms, can live more autonomous life....
This letter shows an effective statistical voice activity detection algorithm based on the integrated bispectrum, which is defined as a cross spectrum between signal and its square inherits ability of higher order statistics to detect signals in noise with many other additional advantages: 1) computation leads significant computational savings, 2) variance estimator same that power estimator. The decision rule formulated terms average likelihood ratio test (LRT) involving successive...