Cisco Technology, Inc.

Eric Chen

Keith Griffin

Wei-Lien Hsu

Nagendra Kumar Nainar

Gonzalo A Salgueiro

Chidambaram Arunachalam

Hugo Latapie

Enzo Fenoglio

Plamen Nedeltchev

Wei Li

Paul G Bright-Thomas

Manikandan Kesavan

Espen Moberg

Asbjørn Therkelsen

Fuling Liu

Nathan Buckles

Cullen Frishman Jennings

Shamim S Pirzada

In one example embodiment, audio characteristics of audio signals are adjusted by a first machine learning model to reduce effects of a facial covering and produce adjusted audio signals. The audio signals correspond to resulting voice signals produced from the facial covering affecting original voice signals. Speech characteristics are predicted for the adjusted audio signals by a second machine learning model. Transformed audio signals corresponding to the original voice signals are produced based on the adjusted audio signals and predicted speech characteristics.

Transforming voice signals to compensate for effects from a facial covering

Systems, methods, and devices are disclosed for training a model. Media data is separated into one or more clusters, each cluster based on a feature from a first model. The media data of each cluster is sampled and, based on an analysis of the sampled media data, an accuracy of the media data of each cluster is determined. The accuracy is associated with the feature from the first model. Based on a subset dataset of the media data being outside a threshold accuracy, the subset dataset is automatically forwarded to a crowd source service. Verification of the subset dataset is received from the crowd source service, and the verified subset dataset is added to the first model.

Continual learning for multi modal systems using crowd sourcing

Systems, methods, and devices are disclosed for detecting an active speaker in a two-way conference. Real time audio in one or more sub band domains are analyzed according to an echo canceller model. Based on the analyzed real time audio, one or more audio metrics are determined from output from an acoustic echo cancellation linear filter. The one or more audio metrics are weighted based on a priority, and a speaker status is determined based on the weighted one or more audio metrics being analyzed according to an active speaker detection model. For an active speaker status, one or more residual echo or noise is removed from the real time audio based on the one or more audio metrics.

Acoustic echo cancellation based sub band domain active speaker detection for audio and video conferencing applications

Chat room content classification, in an online communication environment, where higher applicable chat rooms are prioritized for a user, is provided. First, an chat room service receives chat room content for at least a first chat room and a second chat room. A chat room analyzer can then analyze a characteristic(s) associated with the first chat room and/or the second chat room. Based on the characteristic, the chat room determines that the first chat room is more applicable to the user. Then, a user interface may be presented to the user where the first chat room is prioritized (or ranked) over the second chat room.

Systems and methods to prioritize chat rooms using machine learning

Systems and methods are disclosed for anticipating a video switch to accommodate a new speaker in a video conference comprising a real time video stream captured by a camera local to a first videoconference endpoint is analyzed according to at least one speaker anticipation model. The speaker anticipation model predicts that a new speaker is about to speak. Video of the anticipated new speaker is sent to the conferencing server in response to a request for the video on the anticipated new speaker from the conferencing server. Video of the anticipated new speaker is distributed to at least a second videoconference endpoint.

Speaker anticipation

The subject disclosure relates to systems for managing the deployment and updating of incremental machine learning models across multiple geographic sovereignties. In some aspects, systems of the subject technology are configured to perform operations including: receiving a first machine learning model via a first coordination agent, the first machine learning model based on a first training data set corresponding with a first sovereign region, sending the first machine learning model to a second coordination agent in a second sovereign region, wherein the second sovereign region is different from the first sovereign region, and receiving a second machine learning model from the second coordination agent, wherein the second machine learning model is based on updates to the first machine learning model using a second training data set corresponding with the second sovereign region. Methods and machine-readable media are also provided.

Data sovereignty compliant machine learning

This disclosure relates to solutions for eliminating undesired audio artifacts, such as background noises, on an audio channel. A process for implementing the technology can include receiving a set of audio segments, analyzing the segments using a first ML model to identify a first probability of unwanted background noises in the segments, and if the first probability exceeds a threshold, analyzing the segments using a second ML model to determine a second probability that the one or more background features exist in the segments. In some aspects, the process can include attenuating audio artifacts in the segments, if the second probability exceeds a second threshold. In some implementations, dynamic time stretching and shrinking can be applied to the noise attenuation. Systems and machine-readable media are also provided.

Noise mitigation using machine learning

Systems, methods, and devices are disclosed for detecting an active speaker in a two-way conference. Real time audio in one or more sub band domains are analyzed according to an echo cancellor model. Based on the analyzed real time audio, one or more audio metrics are determined from output from an acoustic echo cancellation linear filter. The one or more audio metrics are weighted based on a priority, and a speaker status is determined based on the weighted one or more audio metrics being analyzed according to an active speaker detection model. For an active speaker status, one or more residual echo or noise is removed from the real time audio based on the one or more audio metrics.

Disclosed is a system, method and computer readable medium enabling collaboration service providers to more accurately predict packet loss, jitter and delay based on current session, historical session and user location parameters. The prediction can be used to forecast the occurrence of poor media quality at the current location and potential future locations.

Media quality prediction for collaboration services

Disclosed is a system and method for forecasting the expected quality of a call. In some examples, a system or method can generate a plurality of scenarios from network metrics, retrieve historical ratings for the network metrics from users, and assign the historical ratings for the network metrics to the plurality of scenarios. The system or method can also filter one or more users based on similarities of the historical ratings for the plurality of scenarios with current network metrics, and forecast an expected call quality based on the historical ratings of the one or more filtered users.

Predictive model for voice/video over IP calls

Disclosed is a system and method for receiving a communication session request and identifying a plurality of available gateways available to handle the communication session request. For each gateway of the plurality of available gateways, gateway metrics of the performance of the gateway are predicted. Based on the predicted gateway metrics, a user rating for the communication session request being handled by the gateway is predicted. Based on the predicted user rating for each gateway, a gateway is selected from the plurality of available gateways.

Growing community of inventors

Palo Alto, CA, United States of America

Eric Chen