Google Inc.

Deepmind Technologies Limited

Thomas Chadwick Walters

Richard Francis Lyon

Ron J Weiss

Douglas Eck

Ryan Michael Rifkin

David A Ross

Jay Yagnik

Oriol Vinyals

Molly Castle Nix

Samy Bengio

Adam L Berenzweig

Jay Michael Ponte

Jason Edward Weston

Aaron Gerard Antonius Van Den Oord

Jason Matthew Toff

Rushabh Ashok Doshi

Sze Chie Lim

Alejandro Luebs

Sally A Goldman

Sam Kvaalen

Robert Steven Glickstein

Matthias Rochus Konrad

Yazhe Li

Christopher Russell LaRosa

Cristina Garbacea

Geremy A Heitz, Iii

Gheorghe Postelnicu

Luciano Sbaiz

Mustafa Suleyman

Luis Carlos Cobo Rus

King Hong Thomas Leung

Thomas Bugnon

Benjamin Kenneth Coppin

Hanna Pasula

Gertjan Pieter Halkes

Timothy Mann

Chia-Yueh Carlton Chu

Martin Szummer

Jean-Francois Crespo

Christopher Larosa

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating discrete latent representations of input audio data. Only the discrete latent representation needs to be transmitted from an encoder system to a decoder system in order for the decoder system to be able to effectively to decode, i.e., reconstruct, the input audio data.

Speech coding using content latent embedding vectors and speaker latent embedding vectors

Generating a playlist may include designating a seed track in an audio library; identifying audio tracks in the audio library having constructs that are within a range of a corresponding construct of the seed track, where the constructs for the audio tracks are derived from frequency representations of the audio tracks, and the corresponding construct for the seed track is derived from a frequency representation of the seed track; and generating the playlist using at least some of the audio tracks that were identified.

Generating a playlist

Systems and techniques for removing a sound recording from an audio recording (e.g., an audio recording embedded in a media file) are presented. The system can include an identification component, a first subtraction component and a second subtraction component. The identification component identifies a sound recording in a mixed audio recording. The first subtraction component determines a local linear transformation of the sound recording and subtracts the local linear transformation of the sound recording from the mixed audio recording to generate a new mixed audio recording. The second subtraction component compares one or more segments of the sound recording with one or more corresponding segments of the new mixed audio recording and reduces a power level of the new mixed audio recording based at least in part on correlation of the one or more corresponding segments with the one or more segments.

System and method for selective removal of audio content from a mixed audio recording

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using a machine learning model that has been trained through reinforcement learning to select a content item. One of the methods includes receiving first data characterizing a first context in which a first content item may be presented to a first user in a presentation environment; and providing the first data as input to a long-term engagement machine learning model, the model having been trained through reinforcement learning to: receive a plurality of inputs, and process each of the plurality of inputs to generate a respective engagement score for each input that represents a predicted, time-adjusted total number of selections by the respective user of future content items presented to the respective user in the presentation environment if the respective content item is presented in the respective context.

Selecting content items using reinforcement learning

Systems and methods facilitating removal of content from audio files are described. A method includes identifying a sound recording in a first audio file, identifying a reference file having at least a defined level of similarity to the sound recording, and processing the first audio file to remove the sound recording and generate a second audio file. In some embodiments, winner-take-all coding and Hough transforms are employed for determining alignment and rate adjustment of the reference file in the first audio file. After alignment, the reference file is filtered in the frequency domain to increase similarity between the reference file and the sound recording. The frequency domain representation (FR) of the filtered version is subtracted from the FR first audio and the result converted to a time representation of the second audio file. In some embodiments, spectral subtraction is also performed to generate a further improved second audio file.

Systems and methods facilitating selective removal of content from a mixed audio recording

Provided content is determined to contain an asset represented by reference content by comparing digital fingerprints of the provided content and the reference content. The fingerprints of the reference content and the provided content are generated using a convolutional neural network (CNN). The CNN is trained using a plurality of frame triplets including an anchor frame representing the reference content, a positive frame which is a transformation of the anchor frame, and a negative frame representing content that is not the reference content. The provided content is determined to contain the asset represented by the reference content based on a similarity measure between the generated fingerprints. If the provided content is determined to contain the asset represented by the reference content, a policy associated with the asset is enforced on the provided content.

Automatic learning of a video matching system

Aspects relate to determining whether a probe media content matches one or more reference media content. The reference media content is classified into a content class. The probe media content could also be classified into a content class. Similarities between the probe media content and the reference media content are identified. A matching score given to the probe media content is weighted based on statistics regarding matches and false-positive rates for the content class of the reference media content. Further, classifiers can be trained on computed audio features and video features and/or video metadata and audio metadata of various media content.

Audio and video matching using a hybrid of fingerprinting and content based classification

Techniques for generating a playlist include: receiving acoustic information, obtaining seed information based on the acoustic information, identifying audio tracks in an audio library based on the seed information, and generating the playlist using at least some of the identified audio tracks.

Generating a playlist based on input acoustic information

Sound representations and winner-take-all codes of auditory spectra are used in the identification of audio content. A transformation component converts a set of sound frames from audio content into a set of spectral slices. A spectral encoder component encodes the spectral slices of auditory spectra into winner-take-all codes with a winner-take-all hash function. An identification component identifies which spectral dimension of a subset of spectral dimensions within a spectral slice has highest spectral value according to the winner-take-all codes. Reference audio content is determined to be similar or matching to the audio content based on the winner-take-all codes.

Sound representation via winner-take-all coding of auditory spectra

A system, method, and computer readable storage medium generates an audio fingerprint for an input audio clip that is robust to differences in key, instrumentation, and other performance variations. The audio fingerprint comprises a sequence of intervalgrams that represent a melody in an audio clip according pitch intervals between different time points in the audio clip. The fingerprint for an input audio clip can be compared to a set of reference fingerprints in a reference database to determine a matching reference audio clip.

Intervalgram representation of audio for melody recognition

Growing community of inventors

San Francisco, CA, United States of America

Thomas Chadwick Walters