Ioannis Dologlou

ASR for emotional speech: Clarifying the issues and enhancing performance

There are multiple reasons to expect that recognising the verbal content of emotional speech will be a difficult problem, and recognition rates reported in the literature are in fact low. Including information about prosody improves recognition rate for emotions simulated by actors, but its relevance to the freer patterns of spontaneous speech is unproven. This paper shows that recognition rate for spontaneous emotionally coloured speech can be improved by using a language model based on increased representation of emotional utterances. The models are derived by adapting an already existing corpus, the British National Corpus (BNC). An emotional lexicon is used to identify emotionally coloured words, and sentences containing these words are recombined with the BNC to form a corpus with a raised proportion of emotional material. Using a language model based on that technique improves recognition rate by about 20%.

Physical interpretation of signal reconstruction from reduced rank matrices

An enhanced version of a signal is obtained after subtraction of a linear combination of rank one matrices from the signal observation matrix. This operation results in a nonToeplitz observation matrix. By averaging the elements of this matrix, a new Toeplitz matrix is produced and a reconstructed signal is generated. This averaging operation is examined, and physical interpretations are given. >

Speech compression based on exact modeling and structured total least norm optimization

We present a new speech coding algorithm, based on an all-pole model of the vocal tract. Whereas current autoregressive (AR) based modeling techniques (e.g. CELP, LPC-10) minimize a prediction error, which is considered to be the input to the all-pole model, our approach determines the closest (in L/sub 2/ norm) signal, which exactly satisfies an all-pole model. Each frame is then encoded by storing the parameters of the complex damped exponentials deduced from the all-pole model and its initial conditions. Decoding is performed by adding the complex damped exponentials based on the transmitted parameters. The new algorithm is demonstrated on a speech signal. The quality is compared with that of a standard coding algorithm at comparable compression ratios, by using the segmental signal-to-noise ratio (SNR).

Pitch detection based on zero-phase filtering

Abstract A new algorithm for the extraction of the fundamental frequency time-varying information is described. The algorithm is based on the iterative use of a linear filter with zero phase and monotonically decreasing frequency response (low pass). The results show that the method is both efficient and robust in noisy environments, providing an estimate for the locations of the closure and opening of the vocal chords.

LPC/SVD analysis of signals with zero modeling error

Abstract The frequency analysis of a signal is reconsidered in this paper. Although the classical approach based on the representation of a signal in terms of complex exponentials has its merit, an alternative approach aiming to represent a signal by a sum of exponential functions is also interesting. An exact solution to this problem can be found at the meeting point of LPC and SVD analysis. Moreover, an optimal representation in the frequency domain can be obtained offering the possibility of an inverse transform when the original signal has to be recomputed. Further insight in the structural analysis of a signal is provided. The rank of a signal is defined and it is shown that it can only be affected by the frequency and the damping factor of an exponential sinusoid.

Decimation and SVD to estimate exponentially damped sinusoids in the presence of noise

A new state-space method for spectral estimation that performs decimation by factor two while it makes use of the full set of data available is presented. The proposed method, called DESE2, is based on singular value decomposition in order to estimate frequency, damping factor, amplitude and phase of exponentially damped sinusoids in the presence of noise. The DESE2 method is compared against some previously proposed methods for spectral estimation that lie among the most promising methods in the field of spectroscopy, where accuracy of parameter estimation is of utmost importance. Experiments performed on a typical simulated NMR signal prove the new method to be more robust, especially for low signal to noise ratio. The new method outperforms the other two not only by presenting lower failure rates but also by incorporating enhanced discriminative analysis while at the same time it benefits from the use of the full data set.

Making assistive reading tools user friendly: a new platform for Greek dyslexic students empowered by automatic speech recognition

This work presents our effort to incorporate a state of the art speech recognition engine into a new platform for assistive reading for improving reading ability of Greek dyslexic students. This platform was developed in the framework of the Agent-DYSL, IST project, and facilitates dyslexic children in learning to read fluently. Unlike previously presented approaches, the aim of the system is not only to enable access to the reading materials within an inclusive learning system but to promote the development of reading skills by adjusting and adapting in the light of feedback to the system. The idea is to improve speech recognition performance so that gradually increase the reading capabilities of the user, gradually diminish the assistance provided, till he is able to read as a non-dyslexic reader. The evaluation results show that both learners’ reading pace and learners’ reading accuracy were increased.

Signal enhancement for continuous speech recognition

This paper presents a comparison between two parametric methods for Signal Enhancement in order to address the problem of robust Automatic Speech Recognition (ASR). An SVD-based technique (ISE) and a non-linear spectral subtraction method (NSS), have been evaluated by means of the Continuous Speech Recognition system that is used in the ERMIS project. The input signal is corrupted with coloured noise with variable signal-to-noise ratio. It was found that fine-tuning of the various parameters of the enhancement techniques is crucial for efficient optimisation of their performance. Both methods provide significant improvement of the speech recogniser performance in the presence of coloured noise, with the NSS method being slightly better.

Spectral estimation based on the eigenanalysis of companion-like matrices

This paper presents a new state-space method for spectral estimation based on a companion matrix technique in order to estimate frequency, damping factor, amplitude and phase of exponential sinusoids. The new method, called CSE (companion matrix based spectral estimation), is compared against a previously proposed method called HTLS which is based on the use of total least squares. The latter lies among the most promising methods in the field of spectroscopy where accuracy of parameter estimation is of utmost importance. Experiments performed on a simulated NMR signal prove CSE to be more robust in terms of failure rate, especially for low signal to noise ratio.

A New Decimative Spectral Estimation Method with Unconstrained Model Order and Decimation Factor

This paper presents a new state-space method for spectral estimation that performs decimation by any factor D while it imposes no constraints to the model order with respect to D. The new method, called DESED, as well as its Total Least Squares version called DESED_TLS, makes use of the full data set available and is based on SVD in order to estimate frequency, damping factor, amplitude and phase of exponential sinusoids.