Download A non-linear technique for room impulse response estimation Most techniques used to estimate the transfer function (or impulse response) of an acoustical space operate along similar principles. A known, broadband signal is transmitted at one point in the room whilst being simultaneously recorded at another. A matched-filter is then used to compress the energy in the transmission waveform in time, forming an approximate impulse response. Finally, equalisation filtering is used to remove any colouration and phase distortion caused by the non-uniform energy-spectrum of the transmission and/or the non-ideal response of the loudspeaker/microphone combination. In this paper, the limitations of this conventional technique will be highlighted, especially when using low-cost equipment. An alternative, non-linear deconvolution technique is proposed which will be shown to give superior performance when using non-ideal equipment.
Download PAEDB: A Synthetic Primary-Ambient Dataset Generation Pipeline for Automatic Upmixing Using Deep Neural Networks Automatic blind upmixing aims to convert audio from a smaller channel format (e.g. mono or stereo) into a multichannel format using estimates of direct and diffuse spatial statistics within the signal. Current approaches rely on primary-ambient extraction (PAE) algorithms, which lack real-world context through limited processing windows. Deep learning music source separation (MSS) models have been applied in voice-primary-ambient extraction (VPA) upmixing systems for handling direct components, but still rely on DSP methods of surround channel generation. This work further investigates utilizing source separation within VPA upmixing, focusing specifically on the task of stereo decorrelation and ambience extraction for 5.1 surround. We also release PAEDB (Primary–Ambient Extraction Dataset), a high-quality music dataset derived from MUSDB18-HQ and MoisesDB, comprising 1,809 primary–ambient stem pairs totaling over 550 hours of audio. The performance of selected DNNs trained on PAEDB is then evaluated using signal metrics and a listening study. Our findings indicate that DNNs can effectively model the behavior of PAE algorithms, establishing PAEDB as a strong foundation for ML upmixing systems and underscoring the need for higher-quality multichannel data to advance beyond conventional methods.