Energy-Based Position Estimation of Microphones and Speakers for Ad Hoc Microphone Arrays

Abstract

We present a novel energy-based algorithm to estimate the positions of microphones and speakers in an ad hoc microphone array setting. Compared to traditional time-of-flight based approaches, energy-based approach has the advantage that it does not require accurate time synchronization. This property is particularly useful for ad hoc microphone arrays because highly accurate synchronization across microphones may be difficult to obtain since these microphones usually belong to different devices. This new algorithm extends our previous energy-based position estimation algorithm [1] in that it does not assume the speakers are in the same positions as their corresponding microphones. In fact, our new algorithm estimates both the microphones and speakers simultaneously. Experiment results are shown to demonstrate its performance improvement over the previous approach in [1], and evaluate its robustness against time synchronization errors.

Whistle: Synchronization-Free TDOA for Localization

“…An energy-based localization system is proposed in [4], in which the positions of speakers are estimated using an ad hoc microphone array setting and the maximum likelihood estimation method. In [21], a set of methods based on energy measurement are developed, compared, and contrasted.…”

“…Existing work basically use three types of measurements of acoustic signals: TDOA [24], [19], [32], [9], DOA (direction of arrival) [22], [14], [12], [15] and received signal energy [25], [26], [4], [21], [3].…”

Ad-hoc microphone array calibration from partial distance measurements

“…A structured sparse coding approach is proposed in [14] to calibrate the microphones through a greedy procedure using an unknown signal at a given position and constructing an overcomplete dictionary of distant signals projected onto the array manifolds. Chen et al [15] introduced an energy-based method for joint microphone calibration and speaker localization by computing the energy of the signal and formulating a nonlinear optimization problem with a non-convex objective function to perform maximum likelihood estimation of the positions. Their approach requires at least five active sources.…”

TDOA-based self-calibration of dual-microphone arrays

“…This problem has been studied using different types of measurements such as received signal strength (RSS) [3], timeof-arrival (TOA) [4][5][6], and time-difference-of-arrival (TDOA) [7]. Among these, TDOA is a suitable choice for HAS applications because it is less vulnerable to reverberation [4], does not require clock-synchronization between sources and microphones, and A typical scenario of microphone array calibration problem for a binaural HAS.…”