Estimating speech spectra for copy synthesis by linear prediction and by hand

Abstract

Linear prediction is a widely available technique for analyzing acoustic properties of speech, although this method is known to be error-prone. New tests assessed the adequacy of linear prediction estimates by using this method to derive synthesis parameters and testing the intelligibility of the synthetic speech that results. Matched sets of sine-wave sentences were created, one set using uncorrected linear prediction estimates of natural sentences, the other using estimates made by hand. Phoneme restrictions imposed on linguistic properties allowed comparisons between continuous and intermittent voicing, oral or nasal and fricative manner, and unrestricted phonemic variation. Intelligibility tests revealed uniformly good performance with sentences created by handestimation and a minimal decrease in intelligibility with estimation by linear prediction due to manner variation with continuous voicing. Poorer performance was observed when linear prediction estimates were used to produce synthetic versions of phonemically unrestricted sentences, but no similar decline was observed with synthetic sentences produced by hand estimation. The results show a substantial intelligibility cost of reliance on uncorrected linear prediction estimates when phonemic variation approaches natural incidence.

Modulation sensitivity in the perceptual organization of speech

“…Estimates of the center frequency and amplitude of vocal resonances were created by hand and used as synthesis parameters for four time-varying sinusoids (see Remez et al, 2011). Temporally distorted versions were created by reversing a brief span of a waveform and composing a new waveform of reversed samples, preserving the original order.…”

Effects of telephone transmission on the performance of formant-trajectory-based forensic voice comparison – Female voices

“…After telephone transmission the measurements for /i/ were approximately 12% lower. On the basis of these 1 Chen et al (2009), Deng et al (2007, Remez et al (2011), and Vallabha and Tuller (2002) discuss some of the difficulties associated with fully-automatic measurement, and Byrne and Foulkes (2004), Duckworth et al (2011), Hillenbrand et al (1995), and Kü nzel (2001) discuss some of the difficulties associated with human-supervised measurement.…”

Across-formant integration and speech intelligibility: Effects of acoustic source properties in the presence and absence of a contralateral interferer

“…In practice, the third-formant contour often corresponded to the fricative formant rather than F3 during phonetic segments with frication; these cases were not treated as errors. Gross errors in automatic estimates of the three formant frequencies were hand-corrected using a graphics tablet; artifacts are not uncommon and manual post-processing of the extracted formant tracks is often necessary (Remez et al, 2011). Amplitude contours corresponding to the corrected formant frequencies were extracted automatically from the stimulus spectrograms; these contours were used to generate synthetic analogues of each sentence.…”