Lorenzo Matassini

Noise reduction for human speech signals by local projections in embedding spaces

A local projective noise-reduction scheme, originally developed for low-dimensional deterministic stationary signals, is successfully applied to human speech. This is possible by exploiting properties of the speech signal which mimic structure exhibited by deterministic chaotic systems. In high-dimensional embedding spaces, the strong nonstationarity is resolved as a sequence of different dynamical regimes of moderate complexity. This filtering technique does not make use of the spectral contents of the signal and is superior to the Ephraim-Malah adaptive filter.

Analysis of vocal disorders in a feature space

This paper provides a way to classify vocal disorders for clinical applications, thanks to the idea of geometric signal separation in a feature space. It is well known that the human voice source generates complex signals including subharmonics and toroidal oscillations. Typical chaotic quantities — like the entropy and the dimension of the attractor — together with autocorrelation function, power spectrum and other conventional measures are analysed in order to provide entries for the feature vectors. We report on a successful application of the geometrical signal separation in distinguishing between normal and disordered phonation. Both qualitative and quantitative results are presented.

On financial markets trading

Starting from the observation of the real trading activity, we propose a model of a stockmarket simulating all the typical phases taking place in a stock exchange. We show that there is no need of several classes of agents once one has introduced realistic constraints in order to confine money, time, gain and loss within an appropriate range. The main ingredients are local and global coupling, randomness, Zipf distribution of resources and price formation when inserting an order. The simulation starts with the initial public offer and comprises the broadcasting of news/advertisements and the building of the book, where all the selling and buying orders are stored. The model is able to reproduce fat tails and clustered volatility, the two most significant characteristics of a real stockmarket, being driven by very intuitive parameters.

Method and apparatus for processing noisy sound signals

A method for processing a sound signal y in which redundancy, consisting mainly of almost repetitions of signal profiles, is detected and correlations between the signal profiles are determined within segments of the sound signal. Correlated signal components are allocated to a power component and uncorrelated signal components to a noise component of the sound signal. The correlations between the signal profiles are determined by methods of nonlinear noise reduction in deterministic systems in reconstructed vector spaces based on the time domain.

Learning the optimal trading strategy

Within a realistic model of the stockmarket, we derive the most successful trading strategy. We first identify the agent who has realized the largest percentual gain and then analyze all the operations this trader has performed during the simulation run. We report them in a proper trading space and we extend the model, introducing an additional operator acting with the help of a look up table derived from a clusterization of space. We discuss the robustness of this optimal strategy, its performance and the applicability to real markets.

Software corrections of vocal disorders.

We discuss how vocal disorders can be post-corrected via a simple nonlinear noise reduction scheme. This work is motivated by the need of a better understanding of voice dysfunctions. This would entail a 2-fold advantage for affected patients: Physicians can perform better surgical interventions and on the other hand researchers can try to build up devices that can help to improve voice quality, i.e. in a phone conversation, avoiding any surgical treatment. As a first step, a proper signal classification is performed, through the idea of geometric signal separation in a feature space. Then through the analysis of the different regions populated by the samples coming from healthy people and from patients affected by T1A glottis cancer, one is able to understand which kind of interventions are necessary in order to correct the illness, i.e. to move the corresponding feature vector from the sick region to the healthy one. We discuss such a filter and show its performance.

On the rigid body behavior of foreign exchange markets

We propose a method to characterize the behavior of a foreign exchange market, introducing a mapping from the N assets to a liquid of N particles, which is studied by means of its empirical kinetic properties. We present the resulting rigid body behavior investigating daily foreign exchange market data for USD, DEM and YEN in the period of time from 9 April 1973 to 15 November 1997.

The trading rectangle strategy within book models

We introduce a model of trading where traders interact through the insertion of orders in the book. This matching mechanism is a collection of the activity of agents: They can trade at the market price or place a limit order. The latter is valid until cancelled by the trader; to this end we introduce a threshold in time after which the probability of the order to be removed is strongly increased. There is essentially no source of randomness and all the traders share a common strategy, what we call trading rectangle. Since there are no fundamentalist rules, it is not so important to identify the right moment to enter in the market. Much more effort is required to decide when to sell. The model is able to reproduce many of the complex phenomena manifested in real stock markets, including the positive correlation between bid/ask spreads and volatility.

Filtering of speech signals by over-embedding

Speech signals possess properties which mimic structures exhibited by deterministic dynamical systems. Furthermore they present strong non-stationarities. In this paper it is shown how a local projective noise reduction scheme, originally developed for low-dimensional deterministic chaotic signals, can be successfully applied to human speech, once an over-embedding is considered. The strong inherent non-stationarity, due to pronounced fluctuations of system parameters, is resolved as a sequence of many different dynamical regimes of moderate complexity, each one corresponding to a single phoneme. If some slowness properties in the change of the parameters are ensured, essential aspects of determinism are restored.

ErratumErratum to “On the three body behavior of foreign exchange markets”: [Physica A 308 (2002) 402]

1. One author has been omitted. The correct list of authors reads L. Matassini and G. Cuniberti (same a3liation). It is obvious that the intellectual property of the paper is also to be attributed to L. Matassini and G. Cuniberti. 2. The term “rigid body” is not only wrong but even misleading since an interaction potential among the three currencies has been introduced in the submitted paper. The term should be substituted with “three body” everywhere in the text (including the title). 3. Finally, the data used for the analysis were provided by MGM Sim and not by Deutsche Boerse AG as erroneously reported in the previously submitted paper.