Enhancement of esophageal speech using formant synthesis.

Abstract

The feasibility of using the formant analysis-synthesis approach to replace the voicing sources of esophageal speech was explored. Using inverse-filtered signals extracted from normal speakers provided the voicing sources. Pitch extraction was tested with various pitch extraction methods, and then a computationally simple, band-limited auto-correlation method was chosen. To accomplish stable and practical speech enhancement, the input signal was divided into low-and highfrequency channels, then only the low-frequency channel was processed by the formant analysissynthesis method. A special purpose DSP-hardware unit was designed to perform the proposed analysis-synthesis process in real-time. Subjective evaluation tests (rating scale method) have been made with seven well-trained esophageal speakers and three speech therapists. Results of the subjective test showed that the synthesized speech was significantly improved, especially in cases of

Esophageal Speech Enhancement Based on Statistical Voice Conversion with Gaussian Mixture Models

“…There have been some attempts based on modifications of its acoustic features, e.g., using comb filtering [1] or smoothing of acoustic parameters [2]. Although they have some efficacy in esophageal speech enhancement, it is basically difficult to compensate for the acoustic differences using those simple modification processes since the acoustic features of esophageal speech exhibit quite different properties from those of normal speech.…”

Improving the recognition of pathological voice using the discriminant HLDA transformation

“…Other attempts to enhance the pathological voices, based on the modification of their acoustic features, e.g. by using comb filtering [15], auditory masking [16], and formant synthesis [17], have been proposed. Although these techniques are useful to improve the quality of pathological speech, it is in practice difficult for them to compensate for the acoustic feature differences between pathological and laryngeal speech.…”

A pattern recognition based esophageal speech enhancement system

“…This means that ψ (t) can be used to analyze and then reconstruct a signal without loss of information [19]. That is the functions given by (6) constitute an unconditional basis in L 2 (R) [19]; and then we can estimate the expansion coefficients of an audio signal f(t) by using the scalar product between f(t) and the function ψ(t) with translation τ and scaling factor s as follows [19]:…”

“…The critical bands theory models the basilar membrane operation as a filter bank in which the bandwidth of each filter increases as its central frequency also increases [6,7]. This requirement can be satisfied using the Bark frequency scale that is a logarithmic scale in which the frequency resolution of any section of the basilar membrane is exactly equal to one Bark, regardless of its characteristic frequency.…”

“…These algorithms reduce the ALT produced background noise, improving the intelligibility, although not the quality, of the ALT produce speech. An analysis-synthesis-based method to improve the alaryngeal speech quality was proposed in [6]. In this method, which uses a vocoder-like approach, the speech is firstly digitalized using a sampling frequency equal to 10 kHz.…”