Model-based distributed node clustering and multi-speaker speech presence probability estimation in wireless acoustic sensor networks

Abstract

A great challenge in the wireless acoustic sensor network (WASN) based signal processing is to develop robust speech presence probability (SPP) estimation methods, which can work at each time frame and each frequency band. The knowledge of SPP plays an essential role in speech enhancement and noise estimation. Single channel SPP estimation and centralized multi-channel SPP estimation have been well studied. However, few efforts can be found for the distributed SPP estimation for WASN applications with multiple speakers. Accordingly, this paper presents a distributed model-based SPP estimation method for multi-speaker detection, which does not need any fusion center. A distributed k-means clustering method is first used to cluster the nodes into subnetworks, which target at detecting different speakers. For each node in the subnetwork, the speech and noise power spectral densities (PSD) are estimated locally by using a model-based method, then a distributed SPP estimator is developed in each subnetwork. A distributed consensus method is used to obtain the distributed clustering and the distributed SPP estimation. The results show that the proposed distributed clustering method can assign nodes into subnetworks based on their noisy observations. Moreover, the proposed distributed SPP estimator achieves robust speech detection performance under different noise conditions.

Sampling Rate Offset Estimation and Compensation for Distributed Adaptive Node-Specific Signal Estimation in Wireless Acoustic Sensor Networks

“…) Higher SRO estimation accuracy can be obtained by determining the non-integer value λ[i] that maximises |p i Γ,kq [λ]|, via an interpolation method such as a golden section search in the interval [λ max [i] − 0.5, λ max [i] + 0.5], as proposed in [13], and substituting λ max [i] by λ[i] in (16).…”

“…which avoids the influence of VAD errors on the results. In practice, the VAD obviously needs to be estimated from the microphone signals [15], [16]. The clock of node 1 is set as the reference, with f s,1 = 16 kHz.…”

Distributed Combined Acoustic Echo Cancellation and Noise Reduction in Wireless Acoustic Sensor and Actuator Networks

“…An ideal VAD was used to isolate the influence of VAD errors. In practice VAD information may be shared among the nodes [31], using a speaker-selective VAD [32] and/or estimating the speech presence probability in a distributed fashion [33]. All nodes in Fig.…”

Unsupervised Clustered Federated Learning in Complex Multi-source Acoustic Environments

“…Microphone clustering in ASNs has been previously explored using e.g., coherence-based features [12], [13], eigenvectors [7], divergence of power spectral densities [14] or cepstral features [6], [15]. These clustering solutions and applications, although effective, do not explicitly incorporate privacy considerations and evaluations are confined to shoebox-type room scenarios.…”