Amazon Technologies, Inc.

Xianxian Zhang

Trausti Thor Kristjansson

Philip Ryan Hilmes

Amit Singh Chhetri

Ramya Gopalan

Laurent Andre Ovaert

A system configured to improve double-talk detection. The system inputs microphone signals into an adaptive filter and determines whether double talk is present based on how the adaptive filter adapts to the microphone signals. For example, when an audible sound is detected, the adaptive filter updates the filter coefficients that correspond to a time difference of arrival of the audible sound. Thus, the device may detect single-talk conditions (e.g., a single peak in the filter coefficients) or double-talk conditions (e.g., two peaks in the filter coefficients). In addition, the device may track a location of the local speech or remote speech over time based on the time difference of arrival.

Methods for detecting double-talk

Techniques for improving acoustic echo cancellation are described. Energy levels of audio data received from a microphone and representing near-end audio and reference audio data representing far-end audio are determined. If near-end audio is detected but far-end audio is not detected, a controller turns of or bypasses an acoustic echo cancellation system until far-end audio is again detected, thereby decreasing or eliminating distortion of the near-end audio by the acoustic echo cancellation system.

Acoustic echo cancellation control

A system configured to improve audio processing by adaptively selecting target signals based on current system conditions. For example, a device may select a target signal based on a highest signal quality metric when only the local speech is present (e.g., during near-end single-talk conditions), as this maximizes an amount of energy included in the output audio signal. In contrast, the device may select the target signal based on a lowest signal quality metric when only the remote speech is present (e.g., during far-end single-talk conditions), as this minimizes an amount of energy included in the output audio signal. In addition, the device may track positions of the local speech and the remote speech over time, enabling the device to accurately select the target signal when both local speech and remote speech is present (e.g., during double-talk conditions).

Beam level based adaptive target selection

Techniques for improving acoustic echo cancellation to attenuate an echo signal generated by a wireless loudspeaker are described. A relative position of the wireless loudspeaker is determined when sound from the wireless loudspeaker is determined to be the dominant sound (e.g., during output of a response to a user command or query or during speech by a far-side party during a two-way audio communication). A beam corresponding to the relative position is thereby selected for echo cancellation.

Acoustic echo cancellation using loudspeaker position

A system configured to improve double-talk detection. The system detects when double-talk is present in a voice conversation using two or more speaker models. The system extracts feature data from microphone audio data and compares the feature data to each speaker model. For example, the system may generate a first distance score indicating a likelihood that the feature data corresponds to a far-end speaker model and a second distance score indicating a likelihood that the feature data corresponds to a universal speaker model. The system may determine current system conditions based on the distance scores and may change settings to improve speech quality during the voice conversation. For example, during far-end single-talk the system may aggressively reduce an echo signal, whereas during near-end single-talk and double-talk the system may apply minimal echo cancellation to improve a quality of the local speech.

A device that includes a first beam selection component that selects a portion of the multiple beams prior to performing adaptive noise cancellation, in addition to a second beam selection component that selects a beam after adaptive noise cancellation is performed. The device only performs adaptive noise cancellation on the portion of the multiple beams that are selected, reducing a complexity of performing adaptive noise cancellation. In addition, as the first beam selection component selects from the multiple beams before adaptive noise cancellation, the quality of the signals are not reduced and/or desired signals are not suppressed, resulting in the first beam selection component increasing a likelihood that the second beam selection component selects the desired beam. The first beam selection component may use a different algorithm and/or selection criteria than the second beam selection component in order to select beams associated with desired signals such as speech.

Growing community of inventors

Santa Clara, CA, United States of America

Xianxian Zhang