Tridimensional assessment of adductor spasmodic dysphonia pre- and post-treatment with Botulinum toxin.

Spasmodic dysphonia voices form, in the same way as substitution voices, a particular category of dysphonia that seems not suited for a standardized basic multidimensional assessment protocol, like the one proposed by the European Laryngological Society. Thirty-three exhaustive analyses were performed on voices of 19 patients diagnosed with adductor spasmodic dysphonia (SD), before and after treatment with Botulinum toxin. The speech material consisted of 40 short sentences phonetically selected for constant voicing. Seven perceptual parameters (traditional and dedicated) were blindly rated by a panel of experienced clinicians. Nine acoustic measures (mainly based on voicing evidence and periodicity) were achieved by a special analysis program suited for strongly irregular signals and validated with synthesized deviant voices. Patients also filled in a VHI-questionnaire. Significant improvement is shown by all three approaches. The traditional GRB perceptual parameters appear to be adequate for these patients. Conversely, the special acoustic analysis program is successful in objectivating the improved regularity of vocal fold vibration: the basic jitter remains the most valuable parameter, when reliably quantified. The VHI is well suited for the voice-related quality of life. Nevertheless, when considering pre-therapy and post-therapy changes, the current study illustrates a complete lack of correlation between the perceptual, acoustic, and self-assessment dimensions. Assessment of SD-voices needs to be tridimensional.

Voicing quantification is more relevant than period perturbation in substitution voices: an advanced acoustical study.

Quality of substitution voicing-i.e., phonation with a voice that is not generated by the vibration of two vocal folds-cannot be adequately evaluated with routinely used software for acoustic voice analysis that is aimed at 'common' dysphonias and nearly periodic voice signals. The AMPEX analysis program (Van Immerseel and Martens) has been shown previously to be able to detect periodicity in irregular signals with background noise, and to be suited for running speech. The validity of this analysis program is first tested using realistic synthesized voice signals with known levels of cycle-to-cycle perturbations and additive noise. Second, exhaustive acoustic analysis is performed of the voices of 116 patients surgically treated for advanced laryngeal cancer and recorded in seven European academic centers. All of them read out a short phonetically balanced passage. Patients were divided into six groups according to the oscillating structures they used to phonate. Results show that features related to quantification of voicing enable a distinction between the different groups, while the features reporting F(0)-instability fail to do so. Acoustic evaluation of voice quality in substitution voices thus best relies upon voicing quantification.

Perceptual evaluation of substitution voices: development and evaluation of the (I)INFVo rating scale.

Substitution voicing cannot be evaluated accurately by the GRBAS perceptual rating scale, and there is a need for a valuable alternative. Therefore, we developed and tried out a perceptual rating scale, consisting of five new parameters: impression, intelligibility, noise, fluency and voicing, each to be scored between 0 (very bad score) to 10 (very good score for a substitution voice). In analogy to the GRBAS scale, they are then converted to deviance scores ranging from 0 (similar to good substitution voicing) to 3 (very deviant from good substitution voicing). Inter-individual agreement measured in a set of 24 semi-professional jury members seemed to be moderate for all parameters. Mean figures of 0.52, 0.51, 0.46, 0.53 and 0.46 are obtained for the parameters impression, intelligibility, noise, fluency and voicing, respectively. Because a high correlation exists between the first two parameters (0.917) and relying on the correlation figures between the two "I"s and the other parameters (correlation values for "impression" vary from 0.79-0.86; values for "intelligibility" range from 0.74-0.83), we suggest to discard the parameter impression, which turns the actual IINFVo scale into INFVo. The proposed (I)INFVo perceptual rating scale seems promising for the assessment of substitution voicing. Eventual improvements and practical proposals are discussed.

[Perceptive evaluation of substitution voices: the I(I) NFVo rating scale]. / Evaluation perceptive des voix de substitution: l'echelle I(I)NFVo.

OBJECTIVES: Evaluation of a perceptual specific rating scale for specific severe non laryngeral dysphonia. MATERIAL AND METHOD: 113 speech samples from substitution voices were scored perceptually according to the IINFVo scale: overall quality impression (I), impression of Intelligibility (I), additive and unnecessary noise (N), speech fluency (F) and presence of voiced segments (Vo). Each parameter was scored on a visual analogue scale from 0 (minimally deviant) to 10 (maximally deviant substitution voicing). These samples were presented to semi-professional jury-members (second grade speech therapy students) and professional jury-members (phoniatricians and speech therapists, specialised in the oncological field). RESULTS: Interindividual agreement between semi-professionals was moderate (0.57-0.68). Interindividual agreement between professionals was higher (0.82-0.86). These figures are similar or even better compared to the classical perceptual evaluation scale for laryngeal speech (GRBAS). CONCLUSION: Our study suggests that the I(I)NFVo rating scale is suitable for perceptual evaluation of substitution voicing and consistency results are comparable with the classical perceptual rating scale (GRBAS).

Detection of spike and wave discharges in the cortical EEG of genetic absence epilepsy rats from Strasbourg.

Genetic absence epilepsy rats from Strasbourg (GAERS) are a strain of Wistar rats in which all animals present spontaneous occurrence of spike and wave discharges (SWD) in the cortical electroencephalogram (EEG). In this paper, we present a method for the detection of SWD, based on the key observation that SWD are quasi-periodic signals. A spectral-comb based analysis method is used to extract the fundamental frequency and the percentage of energy explained by the harmonic spectral components is subsequently used as a detection parameter. It is shown that a maximum sensitivity and specificity of up to 96 per cent can be achieved. We also compared the performance of this method with the methods presented in the literature and conclude that the surplus value of the novel detection method lies in the higher specificity that can be obtained in the analysis of long-term EEG fragments, which are contaminated by artefacts and contain large portions of slow-wave sleep.

An equalized error backpropagation algorithm for the on-line training of multilayer perceptrons.

The error backpropagation (EBP) training of a multilayer perceptron (MLP) may require a very large number of training epochs. Although the training time can usually be reduced considerably by adopting an on-line training paradigm, it can still be excessive when large networks have to be trained on lots of data. In this paper, a new on-line training algorithm is presented. It is called equalized EBP (EEBP), and it offers improved accuracy, speed, and robustness against badly scaled inputs. A major characteristic of EEBP is its utilization of weight specific learning rates whose relative magnitudes are derived from a priori computable properties of the network and the training data.

On the initialization and optimization of multilayer perceptrons.

Multilayer perceptrons are now widely used for pattern recognition, although the training remains a time consuming procedure often converging toward a local optimum. Moreover, as the optimum network size and topology are usually unknown, the search of this optimum requires a lot of networks to be trained. In this paper the authors propose a method for properly initializing the parameters (weights) of a two-layer perceptron, and for identifying (without the need for any error-backpropagation training) the most suitable network size and topology for solving the problem under investigation. The initialized network can then be optimized by means of the standard error-backpropagation (EBP) algorithm. The authors' method is applicable to any two-layer perceptron comprising concentric as well as squashing units on its hidden layer. The output units are restricted to squashing units, but direct connections from the input to the output layer are also accommodated. To illustrate the power of the method, results obtained for different classification tasks are compared to similar results obtained using a traditional error-backpropagation training starting from a random initial state.

Pitch and voiced/unvoiced determination with an auditory model.

In this paper, an accurate pitch and voiced/unvoiced determination algorithm for speech analysis is described. The algorithm is called AMPEX (auditory model-based pitch extractor) and it performs a temporal analysis of the outputs emerging from a new auditory model. However, in spite of its use of an auditory model, AMPEX should not be regarded as a substitute for any psychophysical theory of human auditory pitch perception. What is mainly described is the design of a computationally efficient auditory model, the perceptually motivated determination of the model parameters, the conception of a reliable pitch extractor for speech analysis, and the elaboration of an experimental procedure for evaluating the performance of such a pitch extractor. In the course of the evaluation experiment several kinds of speech stimuli including clean speech, bandpass-filtered speech, and noisy speech were presented to three different pitch extractors. The experimental results clearly indicate that AMPEX outperforms the best algorithms available.

Conducting nonadaptive psychoacoustic or biomedical tests by means of an interactive program.

The program described in this paper was originally designed to generate the (synthetic) stimuli required for a large class of psychoacoustic tests. In our opinion, however, it may also be of interest in other fields of biomedical information processing. The applied computational techniques are standard in the field of digital signal processing, but the program is particular in its way of presenting signals, executing window operations, and describing the decision blocks to be presented during a test. Each decision block is a succession of stimuli and pauses, presented to obtain one judgement from the subject.

A new theory for multitone masking.

In order to explain masked thresholds of a tonal probe centered between two tone maskers, Green [J. Acoust. Soc. Am. 37, 802-813 (1965)] introduced his famous energy-detection model. We argue that the model is in some respects inadequate and inconsistent with measurements of roughness, at least one psychological correlate of temporal fluctuations. Furthermore, it requires the assumption that tones interact over frequency distances considerably beyond a critical bandwidth. Therefore we propose an alternative theory that allows reconciliation of Green's measurements with the commonly accepted critical-band hypothesis. Our model is based on the hypothesis that the auditory system is capable of performing a spectral analysis of the time pattern of the waveform envelope. Important effects of probe type (noise, tone, or complex) and phase between masker and probe seem easily explained by the model. Also, the differences between two- and four-tone masking are successfully predicted. Finally, as a byproduct, our theory offers and explanation for the just noticeable amplitude modulation (AM) and frequency modulation (FM) of a sinusoidal carrier.