Shape parameter estimate for a glottal model without time position

From a recorded speech signal, we propose to estimate a shape parameter of a glottal model without estimating his time position. Indeed, the literature usually propose to estimate the time position first (ex. by detecting Glottal Closure Instants). The vocal-tract filter estimate is expressed as a minimum-phase envelope estimation after removing the glottal model and a standard lips radiation model. Since this filter is mainly biased in low frequencies by the glottal model, an estimation method of a shape parameter is proposed. The evaluation of the results of such an estimator is still difficult. Therefore, this estimator is evaluated with synthetic signals. Such an estimate is useful for voice analysis (ex. glottal source estimation), voice transformation and synthesis.