Devices, systems, and methods for distributed voice processing
摘要
說明書
In these and other implementations, the spatial processor may implement linear filtering or related techniques for selectively disabling/enabling microphones in a way that is not constrained by traditional beamforming techniques. For example, traditional beamforming techniques typically require the number of microphone inputs for a beamformer to be fixed (e.g., to six known microphone channels) because these techniques rely on filtering algorithms that are not adaptive or not readily adaptive to an environment. Linear filtering and related techniques, by contrast, implement algorithms and filtering coefficients that can be adapted on the fly, such that, for example, additional or fewer microphone channels can be selectively routed to the respective voice processors 760a, 760b depending on the particular ambient noise in an environment, available processing power, etc. Additional examples of spatial processors and/or associated filters, such as multi-channel Wiener filters, for processing speech, reverberated speech, and noise signals, s(t), x(t), v(t), may be found in, for example, in previously referenced U.S. patent application Ser. No. 15/984,073 and U.S. patent Ser. No. 16/147,710, filed Sep. 29, 2018, titled “Linear Filtering for Noise-Suppressed Speech Detection Via Multiple Network Microphone Devices,” both of which are incorporated by reference herein in their entireties.
