Voice conversion (VC) refers to transforming the speaker characteristics of an utterance without altering its linguistic contents. Many works on voice conversion require to have parallel training data that is highly expensive to acquire. Recently, the cycle-consistent adversarial network (CycleGAN), which does not require parallel training data, has been applied to voice conversion, showing the state-of-the-art performance. The CycleGAN based voice conversion, however, can be used only for a pair of speakers, i.e., one-to-one voice conversion between two speakers. In this paper, we extend the CycleGAN by conditioning the network on speakers. As a result, the proposed method can perform many-to-many voice conversion among multiple speakers using a single generative adversarial network (GAN). Compared to building multiple CycleGANs for each pair of speakers, the proposed method reduces the computational and spatial cost significantly without compromising the sound quality of the converted voice. Experimental results using the VCC2018 corpus confirm the efficiency of the proposed method.

Steered response power phase transform (SRP-PHAT) is a method that is widely used for robust sound source localization (SSL). However, since SRP-PHAT searches over a large number of candidate locations, it is too slow to run in real-time for large-scale microphone array systems. In this paper, we propose a robust two-level search space clustering method to speed-up SRP-PHAT-based SSL. The proposed method divides the candidate locations of the sound source into a set of groups and finds a small number of groups that are likely to contain the maximum power location. By searching within the small number of groups, the computational costs are reduced by 61.8% compared to a previously proposed method without loss of accuracy.

Voice activity detection (VAD) can be used to distinguish human speech from other sounds, and various applications can benefit from VAD-including speech coding and speech recognition. To accurately detect voice activity, the algorithm must take into account the characteristic features of human speech and/or background noise. In many real-life applications, noise frequently occurs in an unexpected manner, and in such situations, it is difficult to determine the characteristics of noise with sufficient accuracy. As a result, robust VAD algorithms that depend less on making correct noise estimates are desirable for real-life applications. Formants are the major spectral peaks of the human voice, and these are highly useful to distinguish vowel sounds. The characteristics of the spectral peaks are such that, these peaks are likely to survive in a signal after severe corruption by noise, and so formants are attractive features for voice activity detection under low signal-to-noise ratio (SNR) conditions. However, it is difficult to accurately extract formants from noisy signals when background noise introduces unrelated spectral peaks. Therefore, this paper proposes a simple formant-based VAD algorithm to overcome the problem of detecting formants under conditions with severe noise. The proposed method achieves a much faster processing time and outperforms standard VAD algorithms under various noise conditions. The proposed method is robust against various types of noise and produces a light computational load, so it is suitable for use in various applications.

To estimate the unknown distortion parameters from input test signals, estimated transcriptions are typically used for unsupervised adaptation. In a low signal to noise ratio (SNR) condition, the transcription estimated by a decoding procedure can be error prone because of the high mismatch between the acoustic models and the input signal. As a result, it can cause performance degradation of the adapted systems. To account for this problem, we propose an unsupervised adaptation method that can adapt the acoustic models without the estimated transcription. Instead, Gaussian mixture models (GMM) and pseudo phoneme models (PPM) are used. Using these models the unknown distortion parameters are estimated based on the vector Taylor series (VTS) model adaptation scheme. On the Aurora2 task, we obtained relative reduction of 5.4% in word error rate (WER).

This paper presents a transformation-based rapid adaptation technique for robust speech recognition using a linear spectral transformation (LST) and a maximum mutual information (MMI) criterion. Previously, a maximum likelihood linear spectral transformation (ML-LST) algorithm was proposed for fast adaptation in unknown environments. Since the MMI estimation method does not require evenly distributed training data and increases the a posteriori probability of the word sequences of the training data, we combine the linear spectral transformation method and the MMI estimation technique in order to achieve extremely rapid adaptation using only one word of adaptation data. The proposed algorithm, called MMI-LST, was implemented using the extended Baum-Welch algorithm and phonetic lattices, and evaluated on the TIMIT and FFMTIMIT corpora. It provides a relative reduction in the speech recognition error rate of 11.1% using only 0.25 s of adaptation data.

In this paper, we propose a transform-based adaptation technique for robust speech recognition in unknown environments. It uses maximum likelihood spectral transform (MLST) algorithm with additive and convolutional noise parameters. Previously many adaptation algorithms have been proposed in the cepstral domain. Though the cepstral domain may be appropriate for the speech recognition, it is difficult to handle environmental noise directly in the cepstral domain. Therefore our approach deals with such noise in the linear spectral domain in which speech is directly affected by the noise. As a result, we can use a small number of noise parameters for fast adaptation. The experiments evaluated on the FFMTIMIT corpus shows promising result with only a small number of adaptation data.

A Maximum Likelihood Spectral Transformation (MLST) technique is used for robust speech recognition under mismatched training and testing conditions. The linear spectral speech feature vectors of testing utterances are transformed such that the likelihood of the utterances is increased after the transformation. The cepstral vectors are computed from the transformed spectra. The function used for the spectral transformation is designed to handle both convolutional and additive noise. Since the function has small number of parameters to be estimated, MLST requires only a few utterances for adaptation. Furthermore, the computation for parameter estimation and spectral transformation can be done efficiently in linear time. Therefore, the MLST is suitable for rapid online adaptation. To evaluate the efficiency of the MLST technique, it has been implemented for unsupervised incremental online adaptation. The system is tested on speaker-phone telephone speech data, and MLST reduces the error rate by 29.5% when used for speaker and environment adaptation.