Jan Larsen

Detection of skin cancer by classification of Raman spectra

Skin lesion classification based on in vitro Raman spectroscopy is approached using a nonlinear neural network classifier. The classification framework is probabilistic and highly automated. The framework includes a feature extraction for Raman spectra and a fully adaptive and robust feedforward neural network classifier. Moreover, classification rules learned by the neural network may be extracted and evaluated for reproducibility, making it possible to explain the class assignment. The classification performance for the present data set, involving 222 cases and five lesion types, was 80.5%/spl plusmn/5.3% correct classification of malignant melanoma, which is similar to that of trained dermatologists based on visual inspection. The skin cancer basal cell carcinoma has a classification rate of 95.8%/spl plusmn/2.7%, which is excellent. The overall classification rate of skin lesions is 94.8%/spl plusmn/3.0%. Spectral regions, which are important for network classification, are demonstrated to reproduce. Small distinctive bands in the spectrum, corresponding to specific lipids and proteins, are shown to hold the discriminating information which the network uses to diagnose skin lesions.

Generalizable patterns in neuroimaging: how many principal components?

Generalization can be defined quantitatively and can be used to assess the performance of principal component analysis (PCA). The generalizability of PCA depends on the number of principal components retained in the analysis. We provide analytic and test set estimates of generalization. We show how the generalization error can be used to select the number of principal components in two analyses of functional magnetic resonance imaging activation sets.

Wind Noise Reduction using Non-Negative Sparse Coding

We introduce a new speaker independent method for reducing wind noise in single-channel recordings of noisy speech. The method is based on non-negative sparse coding and relies on a wind noise dictionary which is estimated from an isolated noise recording. We estimate the parameters of the model and discuss their sensitivity. We then compare the algorithm with the classical spectral subtraction method and the Qualcomm-ICSI-OGI noise reduction method. We optimize the sound quality in terms of signal-to-noise ratio and provide results on a noisy speech recognition task.

Temporal Feature Integration for Music Genre Classification

Temporal feature integration is the process of combining all the feature vectors in a time window into a single feature vector in order to capture the relevant temporal information in the window. The mean and variance along the temporal dimension are often used for temporal feature integration, but they capture neither the temporal dynamics nor dependencies among the individual feature dimensions. Here, a multivariate autoregressive feature model is proposed to solve this problem for music genre classification. This model gives two different feature sets, the diagonal autoregressive (DAR) and multivariate autoregressive (MAR) features which are compared against the baseline mean-variance as well as two other temporal feature integration techniques. Reproducibility in performance ranking of temporal feature integration methods were demonstrated using two data sets with five and eleven music genres, and by using four different classification schemes. The methods were further compared to human performance. The proposed MAR features perform better than the other features at the cost of increased computational complexity.

Propagation of uncertainty in Bayesian kernel models - application to multiple-step ahead forecasting

The object of Bayesian modelling is predictive distribution, which, in a forecasting scenario, enables evaluation of forecasted values and their uncertainties. We focus on reliably estimating the predictive mean and variance of forecasted values using Bayesian kernel based models such as the Gaussian process and the relevance vector machine. We derive novel analytic expressions for the predictive mean and variance for Gaussian kernel shapes under the assumption of a Gaussian input distribution in the static case, and of a recursive Gaussian predictive density in iterative forecasting. The capability of the method is demonstrated for forecasting of time-series and compared to approximate methods.

Two-Microphone Separation of Speech Mixtures

Separation of speech mixtures, often referred to as the cocktail party problem, has been studied for decades. In many source separation tasks, the separation method is limited by the assumption of at least as many sensors as sources. Further, many methods require that the number of signals within the recorded mixtures be known in advance. In many real-world applications, these limitations are too restrictive. We propose a novel method for underdetermined blind source separation using an instantaneous mixing model which assumes closely spaced microphones. Two source separation techniques have been combined, independent component analysis (ICA) and binary time-frequency (T-F) masking. By estimating binary masks from the outputs of an ICA algorithm, it is possible in an iterative way to extract basis speech signals from a convolutive mixture. The basis signals are afterwards improved by grouping similar signals. Using two microphones, we can separate, in principle, an arbitrary number of mixed speech signals. We show separation results for mixtures with as many as seven speech signals under instantaneous conditions. We also show that the proposed method is applicable to segregate speech signals under reverberant conditions, and we compare our proposed method to another state-of-the-art algorithm. The number of source signals is not assumed to be known in advance and it is possible to maintain the extracted signals as stereo signals.

Generalization performance of regularized neural network models

Architecture optimization is a fundamental problem of neural network modeling. The optimal architecture is defined as the one which minimizes the generalization error. This paper addresses estimation of the generalization performance of regularized, complete neural network models. Regularization normally improves the generalization performance by restricting the model complexity. A formula for the optimal weight decay regularizer is derived. A regularized model may be characterized by an effective number of weights (parameters); however, it is demonstrated that no simple definition is possible. A novel estimator of the average generalization error (called FPER) is suggested and compared to the final prediction error (FPE) and generalized prediction error (GPE) estimators. In addition, comparative numerical studies demonstrate the qualities of the suggested estimator.<<ETX>>

On design and evaluation of tapped-delay neural network architectures

Pruning and evaluation of tapped-delay neural networks for the sunspot benchmark series are addressed. It is shown that the generalization ability of the networks can be improved by pruning using the optimal brain damage method of Le Cun, Denker and Solla. A stop criterion for the pruning algorithm is formulated using a modified version of Akaikes final prediction error estimate. With the proposed stop criterion, the pruning scheme is shown to produce successful architectures with a high yield.<<ETX>>

Independent component analysis for understanding multimedia content

Independent component analysis of combined text and image data from Web pages has potential for search and retrieval applications by providing more meaningful and context dependent content. It is demonstrated that ICA of combined text and image features has a synergistic effect, i.e., the retrieval classification rates increase if based on multimedia components relative to single media analysis. For this purpose a simple probabilistic supervised classifier which works from unsupervised ICA features is invoked. In addition, we demonstrate the suggested framework for automatic annotation of descriptive key words to images.

Comparison of PCA and ICA based clutter reduction in GPR systems for anti-personal landmine detection

This paper presents statistical signal processing approaches for clutter reduction in stepped-frequency ground penetrating radar (SF-GPR) data. In particular, we suggest clutter/signal separation techniques based on principal and independent component analysis (PCA/ICA). The approaches are successfully evaluated and compared on a real SF-GPR time-series. Field-test data are acquired using a monostatic S-band rectangular waveguide antenna.