Voice Activity Detection (VAD) is becoming an essential front-end component in various speech processing systems. As those systems are commonly deployed in environments with diverse noise types and low signal-to-noise ratios (SNRs), an effective VAD method should perform robust detection of speech region out of noisy background signals. In this paper, we propose adversarial domain adaptive VAD (ADA-VAD), which is a deep neural network (DNN) based VAD method highly robust to audio samples with various noise types and low SNRs. The proposed method trains DNN models for a VAD task in a supervised manner. Simultaneously, to mitigate the performance degradation due to back-ground noises, the adversarial domain adaptation method is adopted to match the domain discrepancy between noisy and clean audio stream in an unsupervised manner. The results show that ADA-VAD achieves an average of 3.6%p and 7%p higher AUC than models trained with manually extracted features on the AVA-speech dataset and a speech database synthesized with an unseen noise database, respectively.

Most methods for sound field reconstruction and spherical beamforming with spherical microphone arrays are mathematically based on the spherical harmonics expansion. In many cases, this expansion is truncated at a certain order as in higher order ambisonics (HOA). This truncation leads to performance that is independent of the incident direction of the sound waves. On the other hand, mixed-order ambisonic (MOA) schemes that select an appropriate subset of spherical harmonics can improve the performance for horizontal directions at the expense of other directions. This paper proposes an MOA scheme called Periphony-Lattice to improve sound field reconstruction performance for horizontally incident sound waves. The proposed scheme is compared with the previously introduced MOA and HOA schemes in terms of theoretical truncation error and performance in sound field reconstruction and spherical beamforming. Computer simulations and measurements are conducted with a spherical array of 52 microphones with a nonuniform layout. The results show that the proposed MOA scheme has better performance in sound field reconstruction and spherical beamforming for horizontal sound waves than the other schemes for a given number of microphones. This scheme can be applied to other spherical array layouts if the number of microphones is greater than that of the required spherical harmonics coefficients, and may improve the horizontal performance.