A Hybrid Approach to Electrolaryngeal Speech Enhancement Based on Noise Reduction and Statistical Excitation Generation

Abstract

SUMMARYThis paper presents an electrolaryngeal (EL) speech enhancement method capable of significantly improving naturalness of EL speech while causing no degradation in its intelligibility. An electrolarynx is an external device that artificially generates excitation sounds to enable laryngectomees to produce EL speech. Although proficient laryngectomees can produce quite intelligible EL speech, it sounds very unnatural due to the mechanical excitation produced by the device. Moreover, the excitation sounds produced by the device often leak outside, adding to EL speech as noise. To address these issues, there are mainly two conventional approached to EL speech enhancement through either noise reduction or statistical voice conversion (VC). The former approach usually causes no degradation in intelligibility but yields only small improvements in naturalness as the mechanical excitation sounds remain essentially unchanged. On the other hand, the latter approach significantly improves naturalness of EL speech using spectral and excitation parameters of natural voices converted from acoustic parameters of EL speech, but it usually causes degradation in intelligibility owing to errors in conversion. We propose a hybrid approach using a noise reduction method for enhancing spectral parameters and statistical voice conversion method for predicting excitation parameters. Moreover, we further modify the prediction process of the excitation parameters to improve its prediction accuracy and reduce adverse effects caused by unvoiced/voiced prediction errors. The experimental results demonstrate the proposed method yields significant improvements in naturalness compared with EL speech while keeping intelligibility high enough.

A Vibration Control Method of an Electrolarynx Based on Statistical <i>F</i>₀ Pattern Prediction

“…11. As reported in [9], we confirmed that Batch is significantly improved compared with EL by predicting F 0 patterns based on statistical F 0 patterns. For our proposed methods RT and Forthcoming, we achieved that two proposed systems caused no degradation compared with Batch.…”

“…We found that reducing the variability of F 0 patterns such as rapid movements, we achieved to train F 0 patterns with a smaller number of mixture components. Moreover, as reported in [9], we also confirmed that CF 0 brings better performance compared with the original F 0 because continuous sequence makes it possible to consider inter-frame correlation over an utterance. The proposed segmented CF 0 preserves such an improvement relatively well while minimizing degradation of the prediction accuracy.…”

“…Note that after predicting CF 0 patterns over all frames, only silence frames are automatically detected by using waveform power [9].…”

“…To improve naturalness of EL speech, we have proposed several EL speech enhanced methods based on statistical voice conversion techniques [7]- [9]. In these methods, acoustic features of EL speech are converted into those of normal speech using Gaussian mixture models (GMMs) [7]- [9].…”

“…In these methods, acoustic features of EL speech are converted into those of normal speech using Gaussian mixture models (GMMs) [7]- [9]. We have shown that F 0 pattern replacement from the mechanically generated ones into those predicted from the spectral sequence of the EL speech using the GMM significantly improves naturalness of EL speech while preserving its intelligibility [9]. On the other hand, the use of these enhancement methods needs to use a loudspeaker to present the enhanced EL speech.…”