公开(公告)号：US11735177B2

公开(公告)日：2023-08-22

申请号：US16711889

申请日：2019-12-12

Applicant: Silicon Laboratories Inc.

Inventor： Antonio Torrini ,  Ramin Khoini-Poorfard ,  Sebastian Ahmed

IPC: G10L15/22 ,  G10L25/18 ,  G10L15/16 ,  G10L25/78 ,  G10L15/06 ,  G06N3/08 ,  G10L25/24 ,  G06N3/045 ,  G10L15/08

CPC classification number: G10L15/22 ,  G06N3/045 ,  G06N3/08 ,  G10L15/063 ,  G10L15/16 ,  G10L25/18 ,  G10L25/24 ,  G10L25/78 ,  G10L2015/088 ,  G10L2015/223

Abstract: A system and method of keyword spotting using two neural networks is disclosed. The system is in sleep mode most of the time, and wakes up periodically. Upon waking, a limited duration of audio is examined. This may be performed using an auxiliary neural network. If any audio activity is detected in this duration, the system fully wakes and examines a longer duration of audio for keywords. The keyword spotting is also performed by the main neural network, which may be a convolutional neural network (CNN).

公开(公告)号：US20220335260A1

公开(公告)日：2022-10-20

申请号：US17847408

申请日：2022-06-23

Applicant: Silicon Laboratories Inc.

Inventor： Javier Elenes ,  Antonio Torrini

IPC: G06K9/62 ,  G06N3/08

Abstract: In an embodiment, an apparatus includes: a sensor to sense real world information; a digitizer coupled to the sensor to digitize the real world information into digitized information; a signal processor coupled to the digitizer to process the digitized information into an image; a discriminator coupled to the signal processor to determine, based at least in part on the image, whether the real world information comprises an anomaly, where the discriminator is trained via a generative adversarial network; and a controller coupled to the discriminator.

公开(公告)号：US20210390965A1

公开(公告)日：2021-12-16

申请号：US16898806

申请日：2020-06-11

Applicant: Silicon Laboratories Inc.

Inventor： Antonio Torrini ,  Sebastian Ahmed

IPC: G10L19/02 ,  G10L25/18

Abstract: A system and method of recording and transmitting compressed audio signals over a network is disclosed. The end node device first converts the audio signal to a spectrogram, which is commonly used by machine learning algorithms to perform speech recognition. The end node device then compresses the spectrogram prior to transmission. In certain embodiments, the compression is performed using Discrete Cosine Transforms (DCT). Furthermore, in some embodiments, the DCT is performed on the difference between two columns of the spectrogram. Further, in some embodiments, a function that replaces values below a predetermined threshold with zeroes in the Encoded Spectrogram is utilized. These functions may be performed in hardware or software.

公开(公告)号：US20210266790A1

公开(公告)日：2021-08-26

申请号：US17319875

申请日：2021-05-13

Applicant: Silicon Laboratories Inc.

Inventor： Antonio Torrini ,  Joel Kauppo ,  Sauli Johannes Lehtimaki

IPC: H04W28/06 ,  H04W64/00 ,  H04W40/22 ,  H04W4/02

Abstract: A wireless locator that facilitates determining a location of a nearby wireless asset including an antenna array with multiple antennas, at least one wireless transceiver that receives a location signal from the nearby wireless asset and that takes multiple samples from the location signal including a set of samples for each antenna, and a processor that compresses the samples to generate location information associated with the nearby wireless asset. The wireless locator may be part of a wireless location system including multiple wireless locators distributed in the area and a central processing system. Various compression methods are disclosed, including averaging of the samples, bit reduction of the samples, converting the samples to corresponding phase values, and combining corresponding samples of multiple sample supplemental sets. Combinations of the various compression methods are also disclosed.

公开(公告)号：US20210096207A1

公开(公告)日：2021-04-01

申请号：US16587221

申请日：2019-09-30

Applicant: Silicon Laboratories Inc.

Inventor： Antonio Torrini ,  Sebastian Ahmed ,  Joel Kauppo ,  Sauli Johannes Lehtimaki

IPC: G01S3/04 ,  G06N3/08

Abstract: A system and method of determining the angle of arrival or departure using a neural network is disclosed. The system collects a plurality of I and Q samples as a packet containing a constant tone extension is being received. The I and Q samples are used to form I and Q arrays, which are used as the input to the neural network. The neural network produces a first output representative of the azimuth angle and a second output representative of the elevation angle. In certain embodiments, the neural network is capable of detecting a plurality of angles, where, for each angle, there are three outputs, a first output representative of the azimuth angle, a second output representative of the elevation angle and a third output representative of the relative amplitude. In some embodiments, the neural network is configured to determine the carrier frequency offset of an incoming signal as well.