In this paper we present our system for the detection and classification of acoustic scenes and events (DCASE) 2020 Challenge Task 4: Sound event detection and separation in domestic environments. We introduce two new models: the forward-backward convolutional recurrent neural network (FBCRNN) and the tagconditioned convolutional neural network (CNN). The FBCRNN employs two recurrent neural network (RNN) classifiers sharing the same CNN for preprocessing. With one RNN processing a recording in forward direction and the other in backward direction, the two networks are trained to jointly predict audio tags, i.e., weak labels, at each time step within a recording, given that at each time step they have jointly processed the whole recording. The proposed training encourages the classifiers to tag events as soon as possible. Therefore, after training, the networks can be applied to shorter audio segments of, e.g., 200 ms, allowing sound event detection (SED). Further, we propose a tag-conditioned CNN to complement SED. It is trained to predict strong labels while using (predicted) tags, i.e., weak labels, as additional input. For training pseudo strong labels from a FBCRNN ensemble are used. The presented system scored the fourth and third place in the systems and teams rankings, respectively. Subsequent improvements allow our system to even outperform the challenge baseline and winner systems in average by, respectively, 18.0 % and 2.2 % event-based F1-score on the validation set.

The invention of the Variational Autoencoder enables the application of Neural Networks to a wide range of tasks in unsupervised learning, including the field of Acoustic Unit Discovery (AUD). The recently proposed Hidden Markov Model Variational Autoencoder (HMMVAE) allows a joint training of a neural network based feature extractor and a structured prior for the latent space given by a Hidden Markov Model. It has been shown that the HMMVAE significantly outperforms pure GMM-HMM based systems on the AUD task. However, the HMMVAE cannot autonomously infer the number of acoustic units and thus relies on the GMM-HMM system for initialization. This paper introduces the Bayesian Hidden Markov Model Variational Autoencoder (BHMMVAE) which solves these issues by embedding the HMMVAE in a Bayesian framework with a Dirichlet Process Prior for the distribution of the acoustic units, and diagonal or full-covariance Gaussians as emission distributions. Experiments on TIMIT and Xitsonga show that the BHMMVAE is able to autonomously infer a reasonable number of acoustic units, can be initialized without supervision by a GMM-HMM system, achieves computationally efficient stochastic variational inference by using natural gradient descent, and, additionally, improves the AUD performance over the HMMVAE.

No abstract

In this work we address disentanglement of style and content in speech signals. We propose a fully convolutional variational autoencoder employing two encoders: a content encoder and a style encoder. To foster disentanglement, we propose adversarial contrastive predictive coding. This new disentanglement method does neither need parallel data nor any supervision. We show that the proposed technique is capable of separating speaker and content traits into the two different representations and show competitive speakercontent disentanglement performance compared to other unsupervised approaches. We further demonstrate an increased robustness of the content representation against a train-test mismatch compared to spectral features, when used for phone recognition.

This paper presents an approach to voice conversion, which does neither require parallel data nor speaker or phone labels for training. It can convert between speakers which are not in the training set by employing the previously proposed concept of a factorized hierarchical variational autoencoder. Here, linguistic and speaker induced variations are separated upon the notion that content induced variations change at a much shorter time scale, i.e., at the segment level, than speaker induced variations, which vary at the longer utterance level. In this contribution we propose to employ convolutional instead of recurrent network layers in the encoder and decoder blocks, which is shown to achieve better phone recognition accuracy on the latent segment variables at frame-level due to their better temporal resolution. For voice conversion the mean of the utterance variables is replaced with the respective estimated mean of the target speaker. The resulting log-mel spectra of the decoder output are used as local conditions of a WaveNet which is utilized for synthesis of the speech waveforms. Experiments show both good disentanglement properties of the latent space variables, and good voice conversion performance.