arXiv subjects
Publications and source records attributed to Marcos Faundez-Zanuy.
In this paper we propose a nonlinear scalar predictor based on a combination of Multi Layer Perceptron, Radial Basis Functions and Elman networks. This system is applied to speech coding in an ADPCM backward scheme. The combination of this predictors improves the results of one predictor alone. A comparative study of this three neural networks for speech prediction is also presented.
In this paper we evaluate the relevance of the model size for speaker identification. We show that it is possible to improve the identification rates if a different model size is used for each speaker. We also present some criteria for selecting the model size, and a new algorithm that outperforms the classical system with a fixed model size.
In this paper we propose a nonlinear vectorial prediction scheme based on a Multi Layer Perceptron. This system is applied to speech coding in an ADPCM backward scheme. In addition a procedure to obtain a vectorial quantizer is given, in order to achieve a fully vectorial speech encoder. We also present several results with the proposed system
This paper presents an exhaustive study about the robustness of several parameterizations, with a new database specially acquired for the purpose of a speaker recognition application. This database includes the following variations: different recording sessions (including telephonic and microphonic recordings), recording rooms, and languages (it has been obtained from a bilingual set of speakers). This study has been performed with covariance matrices in a text independent speaker verification application. It reveals that the combination of several parameterizations can improve the robustness in all the scenarios.
In the last years there has been a growing interest for nonlinear speech models. Several works have been published revealing the better performance of nonlinear techniques, but little attention has been dedicated to the implementation of the nonlinear model into real applications. This work is focused on the study of the behaviour of a combined linear/nonlinear predictive model based on linear predictive coding (LPC-10) and neural nets, in a speech waveform coder. Our novel scheme obtains an improvement in SEGSNR between 1 and 2.5 dB for an adaptive quantization ranging from 2 to 5 bits.
This paper describes a novel face identification method that combines the eigenfaces theory with the Neural Nets. We use the eigenfaces methodology in order to reduce the dimensionality of the input image, and a neural net classifier that performs the identification process. The method presented recognizes faces in the presence of variations in facial expression, facial details and lighting conditions. A recognition rate of more than 87% has been achieved, while the classical method of Turk and Pentland achieves a 75.5%.
Advanced motion models (4 or 6 parameters) are needed for a good representation of the motion experimented by the different objects contained in a sequence of images. If the image is split in very small blocks, then an accurate description of complex movements can be achieved with only 2 parameters. This alternative implies a large set of vectors per image. We propose a new approach to reduce the number of vectors, using different block sizes as a function of the local characteristics of the image, without increasing the error accepted with the smallest blocks. A second algorithm is proposed for an inter/intraframe coder.
This paper is focused on nonlinear prediction coding, which consists on the prediction of a speech sample based on a nonlinear combination of previous samples. It is known that in the generation of the glottal pulse, the wave equation does not behave linearly [2], [10], and we model these effects by means of a nonlinear prediction of speech based on a parametric neural network model. This work is centred on the neural net weight's quantization and on the compression gain.
In this paper we summarize several applications based on thermal imaging. We emphasize the importance of emissivity adjustment for a proper temperature measurement. A new set of face images acquired at different emissivity values with steps of 0.01 is also presented and will be distributed for free for research purposes. Among the utilities, we can mention: a) the possibility to apply corrections once an image is acquired with a wrong emissivity value and it is not possible to acquire a new one; b) privacy protection in thermal images, which can be obtained with a low emissivity factor, which is still suitable for several applications, but hides the identity of a user; c) image processing for improving temperature detection in scenes containing objects of different emissivity.
When focusing an image, depth of field, aperture and distance from the camera to the object, must be taking into account, both, in visible and in infrared spectrum. Our experiments reveal that in addition, the focusing problem in thermal spectrum is also hardly dependent of the temperature of the object itself (and/or the scene).
This work defines a procedure for collecting naturally induced emotional facial expressions through the vision of movie excerpts with high emotional contents and reports experimental data ascertaining the effects of emotions on memory word recognition tasks. The induced emotional states include the four basic emotions of sadness, disgust, happiness, and surprise, as well as the neutral emotional state. The resulting database contains both thermal and visible emotional facial expressions, portrayed by forty Italian subjects and simultaneously acquired by appropriately synchronizing a thermal and a standard visible camera. Each subject's recording session lasted 45 minutes, allowing for each mode (thermal or visible) to collect a minimum of 2000 facial expressions from which a minimum of 400 were selected as highly expressive of each emotion category. The database is available to the scientific community and can be obtained contacting one of the authors. For this pilot study, it was found that emotions and/or emotion categories do not affect individual performance on memory word recognition tasks and temperature changes in the face or in some regions of it do not discriminate among emotional states.
Face segmentation is a first step for face biometric systems. In this paper we present a face segmentation algorithm for thermographic images. This algorithm is compared with the classic Viola and Jones algorithm used for visible images. Experimental results reveal that, when segmenting a multispectral (visible and thermal) face database, the proposed algorithm is more than 10 times faster, while the accuracy of face segmentation in thermal images is higher than in case of Viola-Jones
In this paper, we present a pressure characterization and normalization procedure for online handwritten acquisition. Normalization process has been tested in biometric recognition experiments (identification and verification) using online signature database MCYT, which consists of the signatures from 330 users. The goal is to analyze the real mismatch scenarios where users are enrolled with one stylus and then, later on, they produce some testing samples using a different stylus model with different pressure response. Experimental results show: 1) a saturation behavior in pressure signal 2) different dynamic ranges in the different stylus studied 3) improved biometric recognition accuracy by means of pressure signal normalization as well as a performance degradation in mismatched conditions 4) interoperability between different stylus can be obtained by means of pressure normalization. Normalization produces an improvement in signature identification rates higher than 7% (absolute value) when compared with mismatched scenarios.
Practical determination of physical recovery after intense exercise is a challenging topic that must include mechanical aspects as well as cognitive ones because most of physical sport activities, as well as professional activities (including brain computer interface-operated systems), require good shape in both of them. This paper presents a new online handwritten database of 20 healthy subjects. The main goal was to study the influence of several physical exercise stimuli in different handwritten tasks and to evaluate the recovery after strenuous exercise. To this aim, they performed different handwritten tasks before and after physical exercise as well as other measurements such as metabolic and mechanical fatigue assessment. Experimental results showed that although a fast mechanical recovery happens and can be measured by lactate concentrations and mechanical fatigue, this is not the case when cognitive effort is required. Handwriting analysis revealed that statistical differences exist on handwriting performance even after lactate concentration and mechanical assessment recovery. Conclusions: This points out a necessity of more recovering time in sport and professional activities than those measured in classic ways.
This study aimed to assess the reliability and validity of the Polar V800 to measure vertical jump height. Twenty-two physically active healthy men (age: 22.89 +- 4.23 years; body mass: 70.74 +- 8.04 kg; height: 1.74 +- 0.76 m) were recruited for the study. The reliability was evaluated by comparing measurements acquired by the Polar V800 in two identical testing sessions one week apart. Validity was assessed by comparing measurements simultaneously obtained using a force platform (gold standard), high-speed camera and the Polar V800 during squat jump (SJ) and countermovement jump (CMJ) tests. In the test-retest reliability, high intraclass correlation coefficients (ICCs) were observed (mean: 0.90, SJ and CMJ) in the Polar V800. There was no significant systematic bias +- random errors (p > 0.05) between test-retest. Low coefficients of variation (<5%) were detected in both jumps in the Polar V800. In the validity assessment, similar jump height was detected among devices (p > 0.05). There was almost perfect agreement between the Polar V800 compared to a force platform for the SJ and CMJ tests (Mean ICCs = 0.95; no systematic bias +- random errors in SJ mean: -0.38 +- 2.10 cm, p > 0.05). Mean ICC between the Polar V800 versus high-speed camera was 0.91 for the SJ and CMJ tests, however, a significant systematic bias +- random error (0.97 +- 2.60 cm; p = 0.01) was detected in CMJ test. The Polar V800 offers valid, compared to force platform, and reliable information about vertical jump height performance in physically active healthy young men.
We compare a wide band sub-band speech coder using ADPCM schemes with linear prediction against the same scheme with nonlinear prediction based on multi-layer perceptrons. Exhaustive results are presented in each band, and the full signal. Our proposed scheme with non-linear neural net prediction outperforms the linear scheme up to 2 dB in SEGSNR. In addition, we propose a simple method based on a non-linearity in order to obtain a synthetic wide band signal from a narrow band signal.
This paper proposes a multi-section vector quantization approach for on-line signature recognition. We have used the MCYT database, which consists of 330 users and 25 skilled forgeries per person performed by 5 different impostors. This database is larger than those typically used in the literature. Nevertheless, we also provide results from the SVC database. Our proposed system outperforms the winner of SVC with a reduced computational requirement, which is around 47 times lower than DTW. In addition, our system improves the database storage requirements due to vector compression, and is more privacy-friendly as it is not possible to recover the original signature using the codebooks. Experimental results with MCYT provide a 99.76% identification rate and 2.46% EER (skilled forgeries and individual threshold). Experimental results with SVC are 100% of identification rate and 0% (individual threshold) and 0.31% (general threshold) when using a two-section VQ approach.
This paper presents a new algorithm for speaker recognition based on the combination between the classical Vector Quantization (VQ) and Covariance Matrix (CM) methods. The combined VQ-CM method improves the identification rates of each method alone, with comparable computational burden. It offers a straightforward procedure to obtain a model similar to GMM with full covariance matrices. Experimental results also show that it is more robust against noise than VQ or CM alone.