公开(公告)号：US20240249728A1

公开(公告)日：2024-07-25

申请号：US18422523

申请日：2024-01-25

Applicant: Pindrop Security, Inc.

Inventor： Elie KHOURY ,  Matthew GARLAND

IPC: G10L17/08 ,  G06N3/04 ,  G06N3/08 ,  G10L15/16 ,  G10L17/02 ,  G10L17/04 ,  G10L17/18 ,  G10L17/22

CPC classification number: G10L17/08 ,  G06N3/04 ,  G06N3/08 ,  G10L15/16 ,  G10L17/02 ,  G10L17/04 ,  G10L17/18 ,  G10L17/22

Abstract: The present invention is directed to a deep neural network (DNN) having a triplet network architecture, which is suitable to perform speaker recognition. In particular, the DNN includes three feed-forward neural networks, which are trained according to a batch process utilizing a cohort set of negative training samples. After each batch of training samples is processed, the DNN may be trained according to a loss function, e.g., utilizing a cosine measure of similarity between respective samples, along with positive and negative margins, to provide a robust representation of voiceprints.

公开(公告)号：US20190304468A1

公开(公告)日：2019-10-03

申请号：US16442368

申请日：2019-06-14

Applicant: PINDROP SECURITY, INC.

Inventor： Elie KHOURY ,  Matthew GARLAND

IPC: G10L17/00 ,  G10L17/08 ,  G10L15/19 ,  H04M1/27 ,  G10L17/24 ,  G10L15/07 ,  G10L17/04 ,  G06N7/00 ,  G10L15/26

Abstract: Utterances of at least two speakers in a speech signal may be distinguished and the associated speaker identified by use of diarization together with automatic speech recognition of identifying words and phrases commonly in the speech signal. The diarization process clusters turns of the conversation while recognized special form phrases and entity names identify the speakers. A trained probabilistic model deduces which entity name(s) correspond to the clusters.

公开(公告)号：US20180226079A1

公开(公告)日：2018-08-09

申请号：US15890967

申请日：2018-02-07

Applicant: PINDROP SECURITY, INC.

Inventor： Elie KHOURY ,  Matthew GARLAND

IPC: G10L17/26 ,  G10L15/26 ,  G06K9/00 ,  H04L29/06

CPC classification number: G10L17/26 ,  G06F21/32 ,  G06K9/00221 ,  G06K9/00885 ,  G06K9/00926 ,  G06K9/6267 ,  G06K2009/00322 ,  G10L15/265 ,  G10L17/04 ,  G10L17/18 ,  G10L25/30 ,  H04L63/0861

Abstract: A score indicating a likelihood that a first subject is the same as a second subject may be calibrated to compensate for aging of the first subject between samples of age-sensitive biometric characteristics. Age of the first subject obtained at a first sample time and age of the second subject obtained at a second sample time may be averaged, and an age approximation may be generated based on at least the age average and an interval between the first and second samples. The age approximation, the interval between the first and second sample times, and an obtained gender of the subject are used to calibrate the likelihood score.

公开(公告)号：US20180082689A1

公开(公告)日：2018-03-22

申请号：US15709290

申请日：2017-09-19

Applicant: PINDROP SECURITY, INC.

Inventor： Elie KHOURY ,  Matthew GARLAND

IPC: G10L17/00 ,  G10L17/04 ,  G10L15/07 ,  G10L15/26 ,  G06N7/00

CPC classification number: G10L17/005 ,  G06N7/005 ,  G10L15/07 ,  G10L15/19 ,  G10L15/26 ,  G10L17/04 ,  G10L17/08 ,  G10L17/24 ,  H04M1/271 ,  H04M2203/40

Abstract: Utterances of at least two speakers in a speech signal may be distinguished and the associated speaker identified by use of diarization together with automatic speech recognition of identifying words and phrases commonly in the speech signal. The diarization process clusters turns of the conversation while recognized special form phrases and entity names identify the speakers. A trained probabilistic model deduces which entity name(s) correspond to the clusters.

公开(公告)号：US20240212688A1

公开(公告)日：2024-06-27

申请号：US18436911

申请日：2024-02-08

Applicant: PINDROP SECURITY, INC.

Inventor： Ellie KHOURY ,  Matthew GARLAND

IPC: G10L17/00 ,  G06N7/01 ,  G10L15/07 ,  G10L15/19 ,  G10L15/26 ,  G10L17/04 ,  G10L17/08 ,  G10L17/24 ,  H04M1/27

CPC classification number: G10L17/00 ,  G06N7/01 ,  G10L15/07 ,  G10L15/19 ,  G10L15/26 ,  G10L17/04 ,  G10L17/08 ,  G10L17/24 ,  H04M1/271 ,  H04M2203/40

Abstract: Utterances of at least two speakers in a speech signal may be distinguished and the associated speaker identified by use of diarization together with automatic speech recognition of identifying words and phrases commonly in the speech signal. The diarization process clusters turns of the conversation while recognized special form phrases and entity names identify the speakers. A trained probabilistic model deduces which entity name(s) correspond to the clusters.

公开(公告)号：US20230326462A1

公开(公告)日：2023-10-12

申请号：US18329138

申请日：2023-06-05

Applicant: Pindrop Security, Inc.

Inventor： Elie KHOURY ,  Matthew GARLAND

IPC: G10L17/00 ,  H04M1/27 ,  G10L17/24 ,  G10L15/19 ,  G10L17/08 ,  G06N7/01 ,  G10L15/07 ,  G10L15/26 ,  G10L17/04

CPC classification number: G10L17/00 ,  H04M1/271 ,  G10L17/24 ,  G10L15/19 ,  G10L17/08 ,  G06N7/01 ,  G10L15/07 ,  G10L15/26 ,  G10L17/04 ,  H04M2203/40

Abstract: Utterances of at least two speakers in a speech signal may be distinguished and the associated speaker identified by use of diarization together with automatic speech recognition of identifying words and phrases commonly in the speech signal. The diarization process clusters turns of the conversation while recognized special form phrases and entity names identify the speakers. A trained probabilistic model deduces which entity name(s) correspond to the clusters.

公开(公告)号：US20230290357A1

公开(公告)日：2023-09-14

申请号：US18321353

申请日：2023-05-22

Applicant: Pindrop Security, Inc.

Inventor： Elie KHOURY ,  Matthew GARLAND

IPC: G10L17/20 ,  G10L17/02 ,  G10L17/04 ,  G10L17/18 ,  G10L19/028

CPC classification number: G10L17/20 ,  G10L17/02 ,  G10L17/04 ,  G10L17/18 ,  G10L19/028

Abstract: A system for generating channel-compensated features of a speech signal includes a channel noise simulator that degrades the speech signal, a feed forward convolutional neural network (CNN) that generates channel-compensated features of the degraded speech signal, and a loss function that computes a difference between the channel-compensated features and handcrafted features for the same raw speech signal. Each loss result may be used to update connection weights of the CNN until a predetermined threshold loss is satisfied, and the CNN may be used as a front-end for a deep neural network (DNN) for speaker recognition/verification. The DNN may include convolutional layers, a bottleneck features layer, multiple fully-connected layers, and an output layer. The bottleneck features may be used to update connection weights of the convolutional layers, and dropout may be applied to the convolutional layers.

公开(公告)号：US20230037232A1

公开(公告)日：2023-02-02

申请号：US17963091

申请日：2022-10-10

Applicant: Pindrop Security, Inc.

Inventor： Elie KHOURY ,  Matthew GARLAND

IPC: G10L17/08 ,  G10L15/16 ,  G10L17/02 ,  G10L17/04 ,  G10L17/22 ,  G06N3/04 ,  G06N3/08 ,  G10L17/18

Abstract: The present invention is directed to a deep neural network (DNN) having a triplet network architecture, which is suitable to perform speaker recognition. In particular, the DNN includes three feed-forward neural networks, which are trained according to a batch process utilizing a cohort set of negative training samples. After each batch of training samples is processed, the DNN may be trained according to a loss function, e.g., utilizing a cosine measure of similarity between respective samples, along with positive and negative margins, to provide a robust representation of voiceprints.

公开(公告)号：US20210082439A1

公开(公告)日：2021-03-18

申请号：US17107496

申请日：2020-11-30

Applicant: PINDROP SECURITY, INC.

Inventor： Elie KHOURY ,  Matthew GARLAND

IPC: G10L17/20 ,  G10L17/02 ,  G10L17/04 ,  G10L17/18 ,  G10L19/028

Abstract: A system for generating channel-compensated features of a speech signal includes a channel noise simulator that degrades the speech signal, a feed forward convolutional neural network (CNN) that generates channel-compensated features of the degraded speech signal, and a loss function that computes a difference between the channel-compensated features and handcrafted features for the same raw speech signal. Each loss result may be used to update connection weights of the CNN until a predetermined threshold loss is satisfied, and the CNN may be used as a front-end for a deep neural network (DNN) for speaker recognition/verification. The DNN may include convolutional layers, a bottleneck features layer, multiple fully-connected layers and an output layer. The bottleneck features may be used to update connection weights of the convolutional layers, and dropout may be applied to the convolutional layers.

公开(公告)号：US20200302939A1

公开(公告)日：2020-09-24

申请号：US16895750

申请日：2020-06-08

Applicant: PINDROP SECURITY, INC.

Inventor： Elie KHOURY ,  Matthew GARLAND

IPC: G10L17/00 ,  H04M1/27 ,  G10L17/24 ,  G10L15/19 ,  G10L17/08 ,  G06N7/00 ,  G10L15/07 ,  G10L15/26 ,  G10L17/04

Abstract: Utterances of at least two speakers in a speech signal may be distinguished and the associated speaker identified by use of diarization together with automatic speech recognition of identifying words and phrases commonly in the speech signal. The diarization process clusters turns of the conversation while recognized special form phrases and entity names identify the speakers. A trained probabilistic model deduces which entity name(s) correspond to the clusters.