Infineon Technologies Ag

Cypress Semiconductor Corporation

Friedrich-Alexander-Universitat Erlangen-Nürnberg

Avik Santra

Ashutosh Baheti

Saverio Trotta

Reinhard-Wolfgang Jungmaier

Raghavendran Vagarappan Ulaganathan

Jagjit Singh Bal

Prachi Vaishnav

Muhammad Arsalan

Thomas Reinhold Stadelmayer

Thomas Finke

Souvik Hazra

Lorenzo Servadei

Neha Baheti

Ashutosh Pandey

Ivan Tsvelykh

Samo Vehovc

Christoph Will

Michael Stephan

Johann Peter Forstner

Dennis Noppeney

Cynthia Kuo

Andreas Dorfner

Christoph Rumpler

Andrea Heinz

Jens Reinstaedt

Youn Joo Kim

Lorenz Ferdinand Wilhelm Weiland

Maher Matta

Aidan Smyth

Niall Lyons

Mandy Barsilai Fernandez

Thomas Gmeinder

Henk Muller

Shyam Nallabolu

Assad Ali

Laurent Remont

Julius Ott

Ivan Pavlov

Baheti Neha

Jens Reinstädt

Johann Peter Hans Forstner

Implementations disclosed describe methods, devices, and systems to perform out-of-distribution and recognition of activities using an inertial measurement unit (IMU) sensor. A method may include receiving motion data by a device from a motion sensor. The method further includes generating image data comprising one or more images based on the motion data. The method further includes determining that a first portion of the image data corresponds to activities outside a classification distribution. The method further includes filtering the image data by removing the first portion from the image and generating filtered image data. The method further includes determining an activity classification, within the classification distribution, based on the filtered image data. The method further includes modifying an operating parameter of the device based on the activity classification.

Out-of-distribution detection and recognition of activities with inertial measurement unit sensor

In an embodiment, a method includes: generating a target displacement signal indicative of a movement of a human target based on raw digital data generated by a millimeter-wave radar sensor; and estimating a vital sign of the human target based on the target displacement signal, where generating the target displacement signal includes: generating target in-phase (I) and quadrature (Q) (I/Q) data associated with the human target based on the raw digital data, classifying the target I/Q data as high quality data or as low quality data using a first neural network, when the target I/Q data is classified as low quality data, discarding the target I/Q data, when the target I/Q data is classified as high quality data, performing ellipse fitting on the target I/Q data to generate compensated I/Q data, and generating the target displacement signal based on the compensated I/Q data.

Radar-based single target vital sensing

In an embodiment, a method of tracking includes predicting a predicted state, in a global coordinate space, of an object based on a state of the object; determining in local coordinates the predicted state; determining a plurality of measurements of the object, in the local coordinates, with first and/or second radar sensors; determining a matching of the predicted state and the plurality of measurements, in the local coordinates, for a matching result; and updating the state of the object based on the matching result. The first and second sensors are arranged along perpendicular lines which intersect at the origin of the global coordinate space.

Position tracking with multiple sensors

In an embodiment, a method to evaluate radar images includes providing a first raw radar image and a second raw radar image and determining, whether a reliability criterion is fulfilled. The method further includes using a first coordinate and a second coordinate output by a trained neural network as an estimate of a position of an object if the reliability criterion is fulfilled, the trained neural network using the first raw radar image and the second raw radar image as an input. The method further includes using a third coordinate and a fourth coordinate output by another radar processing pipeline as the estimate of the position of the object if the reliability criterion is not fulfilled, the radar processing pipeline using the first raw radar image and the second raw radar image as an input.

Method and apparatus to evaluate radar images and radar device

An earphone device includes a housing comprising a top region and a bottom region, an acoustic transducer disposed in the bottom region of the housing, and a radar system disposed in the top region of the housing. The radar system includes a first side and an opposite second side. The radar system is configured to detect a first object located on the first side of the radar system, and detect biometric data from a second object located on the second side of the radar system.

Multifunctional radar systems and methods of operation thereof

In an embodiment, a method includes: obtaining one or more positional time spectrograms of a radar measurement of a scene comprising an object; and based on the one or more positional time spectrograms and based on a feature embedding of a variational auto-encoder neural network, predicting a gesture class of a gesture performed by the object.

Radar-based gesture classification using a variational auto-encoder neural network

In an embodiment, a method includes: receiving first radar data from a millimeter-wave radar sensor; receiving a set of hyperparameters with a radar processing chain; generating a first radar processing output using the radar processing chain based on the first radar data and the set of hyperparameters; updating the set of hyperparameters based on the first radar processing output using a hyperparameter selection neural network; receiving second radar data from the millimeter-wave radar sensor; and generating a second radar processing output using the radar processing chain based on the second radar data and the updated set of hyperparameters.

Scene-adaptive radar

In an embodiment, a method includes: obtaining one or more radar measurement frames, each one of the one or more radar measurement frames including respective data samples acquired by a radar sensor monitoring a scene; for each one of the one or more radar measurement frames, determining a respective 2-D angular intensity map of the scene based on the respective radar measurement frame; and performing a people counting operation based on the one or more 2-D angular intensity maps determined for the one or more radar measurement frames to determine a people count for the scene.

People counting based on radar measurement

In an embodiment, a method includes: receiving raw data from a millimeter-wave radar sensor; generating a first radar-Doppler image based on the raw data; generating a first radar point cloud based on the first radar-Doppler image; using a graph encoder to generate a first graph representation vector indicative of one or more relationships between two or more parts of the target based on the first radar point cloud; generating a first cadence velocity diagram indicative of a periodicity of movement of one or more parts of the target based on the first radar-Doppler image; and classifying an activity of a target based on the first graph representation vector and the first cadence velocity diagram.

Radar-based activity classification

In an embodiment, a method for tracking a target includes: receiving raw data from a millimeter-wave radar, the raw data including a plurality of macro-Doppler frames, each macro-Doppler frame having N chirps, where each macro-Doppler frame stretches over a time interval having a first duration; generating micro-Doppler frames from the plurality of macro-Doppler frames, each micro-Doppler frame including L chirps from M macro-Doppler frames, where each micro-Doppler frame stretches over a time interval having a second duration that is longer than the first duration; detecting one or more moving targets based on the macro-Doppler frames; detecting one or more static targets based on the micro-Doppler frames; and tracking a first target as the target transitions from being detected based on the macro-Doppler frames to being detected based on the micro-Doppler frames.

Radar-based target tracker

In an embodiment, a method includes: obtaining depth data indicative of a time-resolved measurement of a 3-D position of a moveable target in a field-of-view of a depth sensor; and processing the depth data using an interacting multi-model (IMM) tracking algorithm, the IMM tracking algorithm providing, for each one of multiple iterations, tracking data comprising a respective estimate of the 3-D position of the moveable target, wherein the IMM tracking algorithm comprises a first model providing a first output, a second model providing a second output, and a fusion module fusing the first output and the second output to provide the tracking data, wherein the first model models a movement state of the moveable target, and wherein the second model models a rest state of the moveable target.

Interacting multi-model tracking algorithm using rest state model

A system includes a millimeter-wave radar sensor disposed on a circuit board, a plurality of antennas coupled to the millimeter-wave radar sensor and disposed on the circuit board, and a processing circuit coupled to the millimeter-wave radar sensor and disposed on the circuit board. The processing circuit is configured to determine vital signal information based on output from the millimeter-wave radar sensor.

System and method for vital signal sensing using a millimeter-wave radar sensor

Described are techniques for noise-robust and speaker-independent keyword spotting (KWS) in an input audio signal that contains keywords used to activate voice-based human-computer interactions. A KWS system may combine the latent representation generated by a denoising autoencoder (DAE) with audio features extracted from the audio signal using a machine learning approach. The DAE may be a discriminative DAE trained with a quadruplet loss metric learning approach to create a highly-separable latent representation of the audio signal in the audio input feature space. In one aspect, spectral characteristics of the audio signal such as Log-Mel features are combined with the latent representation generated by a quadruplet loss variational DAE (QVDQE) as input to a DNN KWS classifier. The KWS system improves keyword classification accuracy versus using extracted spectral features alone, non-discriminative DAE latent representations alone, or the extracted spectral features combined with the non-discriminative DAE latent representations in a KWS classifier.

Noise robust representations for keyword spotting systems

A wireless multimode system includes: an array of N antenna elements that includes a first portion of M antenna elements and a second portion of L antenna elements; M transmission amplifiers configured to transmit, via the M antenna elements, frames of transmit data, where the frames of transmit data include transmit radar signals and transmit communication signals; M reception amplifiers configured to receive, via the M antenna elements, frames of receive data, where the frames of receive data includes receive communication signals; and L reception amplifiers configured to receive, via the L antenna elements, receive radar signals; and a resource scheduler configured to allocate bandwidth for transmit radar signals and transmit communication signals within the frames of transmit data based on one or more predetermined parameters.

Multimode communication and radar system resource allocation

In an embodiment, a method includes: transmitting a plurality of radar signals using a millimeter-wave radar sensor towards a target; receiving a plurality of reflected radar signals that correspond to the plurality of transmitted radar signals using the millimeter-wave radar; mixing a replica of the plurality of transmitted radar signals with the plurality of received reflected radar signals to generate an intermediate frequency signal; generating raw digital data based on the intermediate frequency signal using an analog-to-digital converter; processing the raw digital data using a constrained L dimensional convolutional layer of a neural network to generate intermediate digital data, where L is a positive integer greater than or equal to 2, and where the neural network includes a plurality of additional layers; and processing the intermediate digital data using the plurality of additional layers to generate information about the target.

Growing community of inventors

Munich, Germany

Avik Santra