Filter and method for informed spatial filtering using multiple instantaneous direction-of-arrivial estimates

Simple SummaryContent extracted from patent full text and abstract with AI.

This patent describes a sophisticated audio filter and method that processes sound captured by multiple microphones in order to produce a cleaner, enhanced audio output. By analyzing where sounds are coming from (the directions-of-arrival or DOA) for each small segment in time and frequency, the filter dynamically generates weights that combine the microphone signals to boost desired sounds (e.g., a specific speaker) and suppress unwanted ones (such as noise or reverberation). The system flexibly supports multiple active sound sources and can quickly adapt to changing acoustic environments for high-quality audio extraction or reproduction.

Use CasesContent extracted from patent full text and abstract with AI.

• Speech enhancement in teleconferencing and communication systems, such as video calls and hearing aids
• Audio recording and broadcast in noisy or reverberant environments (lecture halls, conference rooms, studios)
• Smart home assistants and voice-controlled devices for accurate speaker localization and noise suppression
• Spatial audio capture and reproduction in VR/AR applications or immersive sound systems
• Surveillance and security systems for tracking or isolating audio sources in complex soundscapes
• Binaural or multi-channel sound recording for realistic playback in headphones or multi-speaker setups

BenefitsContent extracted from patent full text and abstract with AI.

• Improved signal clarity by effectively separating desired sounds from noise and reverberation
• Fast adaptation to changing acoustic environments and moving sound sources
• Supports multiple simultaneous speakers or sources, enhancing real-world applicability
• Flexible output: can be tailored for mono, stereo, binaural, or multi-channel reproduction
• Highly customizable spatial filtering, allowing arbitrary directivity patterns
• Efficient and robust: combines statistical and real-time data for optimal performance, leading to better audio quality in challenging acoustic scenarios
• Applicable via software or hardware and can be integrated into existing processing systems

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

A filter (100) for generating an audio output signal, comprising a plurality of audio output signal samples, based on two or more input microphone signals is provided. The audio output signal and the two or more input microphone signals are represented in a time-frequency domain, wherein each of the plurality of audio output signal samples is assigned to a time-frequency bin (( k , n)) of a plurality of time-frequency bins ((k, n )). The filter (100) comprises a weights generator (110) being adapted to receive, for each of the plurality of time-frequency bins (( k , n)), direction-of-arrival information of one or more sound components of one or more sound sources or position information of one or more sound sources, and being adapted to generate weighting information for each of the plurality of time-frequency bins ((k, n)) depending on the direction-of-arrival information of the one or more sound components of the one more sound sources of said time-frequency bin (( k , n )) or depending on the position information of the one or more sound sources of said time-frequency bin (( k , n )). Moreover, the filter comprises an output signal generator (120) for generating the audio output signal by generating for each of the plurality of time-frequency bins (( k , n )) one of the plurality of audio output signal samples, which is assigned to time-frequency bin (( k , n )), depending on the weighting information of said time-frequency bin (( k , n )) and depending on an audio input sample, being assigned to said time-frequency bin (( k, n )), of each of the two or more input microphone signals.