This paper compares the performance of three text-to-speech (TTS) models released from June 2021 to January 2022 in order to establish a baseline for Brazilian Portuguese. Those models were trained using dataset for Brazilian Portuguese. The experimental setup considers tts-portuguese dataset to fine-tune the following TTS models: VITS end-to-end model; glowtts and gradtts acoustic models both using hifigan vocoder. Performance metrics are arranged into objective and subjective metrics. As subjective metrics, the naturalness and intelligibility are measured based on the mean opinion score (MOS). Results shows that gradtts+hifigan model achieved naturalness of 4.07 MOS, close to performance of current commercial models.

This paper presents a robust beamformer for stereo noise reduction in hearing aid applications. The worst-case optimization method was applied to the binaural minimum-variance distortionless-response (BMVDR) beamformer, for providing robustness against parameter estimation inaccuracies. This property is essential in real-world applications, since it is well known that steering-vector and noise-correlation estimation errors may significantly degrade the BMVDR performance. It is shown that its control parameters can be designed as a function of the noisy-speech interaural-level-difference estimate. Simulation experiments for synthetic and real-world noise, as well as for a range of signal to interference ratios (SIR) were performed. For a particular set of fixed control parameters, it is shown that the mean performance of the conventional BMVDR may be improved by 0.76 wideband perceptual evaluation of speech quality (WPESQ). The proposed method is especially effective for input SIRs higher than 10 dB, which is a crucial range for hearing aid users.