Mikko Parviainen

Self-localization of wireless acoustic sensors in meeting rooms

This paper presents a passive acoustic self-localization and synchronization system, which estimates the positions of wireless acoustic sensors utilizing the signals emitted by the persons present in the same room. The system is designed to utilize common off-the-shelf devices such as mobile phones. Once devices are self-localized and synchronized, the system could be utilized by traditional array processing methods. The proposed calibration system is evaluated with real recordings from meeting scenarios. The proposed system builds on earlier work with the added contribution of this work is i) increasing the accuracy of positioning, and ii) introduction data-driven data association. The results show that improvement over the existing methods in all tested recordings with 10 smartphones.

Moving sound source localization in large areas

This paper presents a sound source localization method that considers the limited propagation speed of sound. The localization system is shown to be independent of source movement. Information from a sound source is received with a propagation delay relative to the receiver distance. Combining information from spatially separate sensor stations utilizing propagation delays offers a realistic approach to passive sound source localization. The discussed bearings-only localization method uses the direction of arrival information received from a network of sensor stations. It is shown that incorporating a simple propagation model into large area localization results in better performance than neglecting it. The increase of computational capacity has made the implementation of the discussed non-analytical method feasible.

Robust Speaker Localization in Meeting Room Domain

Speech signal is perturbed in real environments due to room acoustics and noise sources. The speech wavefront is received by a microphone array, which is used to determine the wavefront direction of arrival (DOA). DOA estimates from several spatially separated arrays are used to locate the speaker. Array channels may sometimes be corrupted by noise. In these cases, a DOA estimate may differ from the actual direction of the speaker. Using such estimates in localization could deteriorate the position estimate. A criterion is presented for DOA exclusion. A subset of DOA estimates is chosen that minimizes this criterion. The method is numerically shown to improve localization robustness. Recordings from a real meeting room are studied.

A spatiotemporal approach to passive sound source localization

The delay between the event and the observation is a fundamental problem in acoustic localization. This fact is accounted for in the presented localization approach as an additional stage in the applied model. This work concentrates on passive acoustic source localization from a spatiotemporal viewpoint. The direction of arrival (DOA) estimates are measured over time at spatially separated sensor stations, e.g. microphone arrays. The DOA estimates are combined to produce source likelihood in such a manner that the propagation delays are compensated. This paper focuses on incorporating the propagation delay model rather than studying the properties of the localization algorithm itself. The propagation delay correction is substantial in large area surveillance.

TUT acoustic source tracking system 2006

This paper documents the acoustic source tracking system developed by TUT for the 2006 CLEAR evaluation campaign. The described system performs 3-D single person tracking based on audio data received from multiple spatially separated microphone arrays. The evaluation focuses on meeting room domain. The system consists of four distinct stages. First stage is time delay estimation (TDE) between microphone pairs inside each array. Based on the TDE, direction of arrival (DOA) vectors are calculated for each array using a confidence metric. Source localization is done by using a selected combination of DOA estimates. The location estimate is tracked using a particle filter to reduce noise. The system is capable of locating a speaker 72% of the time with an average accuracy of 25 cm.

Spatiotemporal approach for passive sound source localization - real-world experiments

Summary form only given. An approach to consider propagation time delay in sound source localization is tested using recorded data. Sound sources can be localized with a network of acoustic sensor stations. Each sensor station is a subsystem calculating direction of arrival of a dominant sound source in their audibility range. The finite propagation speed from a sound source to a sensor station has to be considered in localization. In a network based localization, the reception time of acoustic events is usually different at each sensor station. The difference has to be considered while combining the information of spatially separate sensor stations. If the difference is neglected, it may result in deteriorated location estimation. The results obtained with recorded data show that the spatiotemporal approach is able to take into account the propagation delay and thus compensate the fact that the sensor stations receive the information at different times.

Obtaining an optimal set of head-related transfer functions with a small amount of measurements

This article presents a method to obtain personalized Head-Related Transfer Functions (HRTFs) for creating virtual soundscapes based on small amount of measurements. The best matching set of HRTFs are selected among the entries from publicly available databases. The proposed method is evaluated using a listening test where subjects assess the audio samples created using the best matching set of HRTFs against a randomly chosen set of HRTFs from the same location. The listening test indicates that subjects prefer the proposed method over random set of HRTFs.

Noise-robust detection of whispering in telephone calls using deep neural networks

Detection of whispered speech in the presence of high levels of background noise has applications in fraudulent behaviour recognition. For instance, it can serve as an indicator of possible insider trading. We propose a deep neural network (DNN)-based whispering detection system, which operates on both magnitude and phase features, including the group delay feature from all-pole models (APGD). We show that the APGD feature outperforms the conventional ones. Trained and evaluated on the collected diverse dataset of whispered and normal speech with emulated phone line distortions and significant amounts of added background noise, the proposed system performs with accuracies as high as 91.8%.

Robust self-localization solution for meeting room environments

This work presents a robust scheme for automatically discovering the coordinates of nodes in an acoustic sensor network. The system is based on utilizing measured Time Difference of Arrival (TDOA) information. The focus is on how to increase the robustness of the system in configurations which render the problem ill-posed. Truncated Singular Value Composition and Tikhonov Regularization are analyzed to cover an ill-posed problem with data produced by an acoustic sensor network. The initial experiments published in this work show that the robustness of such networks can be increased if either of the tested methods is used.