Abstract
The paper discusses a measurement approach for the room impulse response (RIR), which is insensitive to the nonlinearities that affect the measurement instruments. The approach employs as measurement signals the perfect periodic sequences for Wiener nonlinear (WN) filters. Perfect periodic sequences (PPSs) are periodic sequences that guarantee the perfect orthogonality of a filter basis functions over a period. The PPSs for WN filters are appealing for RIR measurement, since their sample distribution is almost Gaussian and provides a low excitation to the highest amplitudes. RIR measurement using PPSs for WN filters is studied and its advantages and limitations are discussed. The derivation of PPSs for WN filters suitable for RIR measurement is detailed. Limitations in the identification given by the underestimation of RIR memory, order of nonlinearity, and effect of measurement noise are analysed and estimated. Finally, experimental results, which involve both simulations using signals affected by real nonlinear devices and real RIR measurements in the presence of nonlinearities, compare the proposed approach with the ones that are based on PPSs for Legendre nonlinear filter, maximal length sequences, and exponential sweeps.
Highlights
Measuring the room impulse response (RIR) is a basic operation for acoustics and audio signal processing.It is needed for analyzing and characterizing the impulse response, estimating parameters, like reverberation time, early decay time, center-of-gravity time, clarity, definition, warmth, brilliance, interaural cross-correlation, lateral energy fraction, etc. [1,2]
We extend the approach of [36] and consider the robust RIR measurement using Perfect periodic sequences (PPSs) for Wiener nonlinear (WN) filters
The results obtained with the proposed approach are compared with those obtained with the PPSs for Legendre nonlinear (LN) filters [35,36], the maximal length sequence (MLS) [14], and the exponential sweeps technique [21,22]
Summary
Measuring the room impulse response (RIR) is a basic operation for acoustics and audio signal processing. The first-order kernel of the LN filter can be estimated by computing the cross-correlation between the output and PSS input signals Another family of polynomial filters having orthogonal basis functions is the Wiener nonlinear (WN) filters, which derive from the truncation of the Wiener nonlinear series. The paper provides novel experimental results that compare the proposed approach with competing approaches, i.e., MLSs, exponential sweeps, PPSs for LN filters. We consider both simulations that invove recorded signals affected by real nonlinear devices and convolved with a measured RIR, and real measurements performed in a room with professional equipment.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have