University of Illinois

Haitham Al Hassanieh

Romit Roy Choudhury

Nirupam Roy

Junfeng Guan

Suraj S Jog

Seyedsohrab Madani

Saurabh Gupta

A method includes receiving data including a plurality of data items, each data item of the plurality of data items including a three-dimensional (3D) radar heat map of an object and a corresponding two-dimensional (2D) image of the object captured by a stereo camera, inputting the training dataset into a machine learning model including a neural network (NN) that generates, from the 3D radar heat map, a 2D depth map for the object and outputs a probability that the 2D depth map is the corresponding 2D image of the object, and training the machine learning model based on a training dataset to generate a trained machine learning model that iteratively learns to generate an updated 2D depth map that approximates the corresponding 2D image. The training dataset includes the plurality of data items, the 2D depth map and the probability.

Neural network-based millimeter-wave imaging system

A voice enabled device includes a transducer to capture multiple inaudible signals received from multiple ultrasonic speakers and audio recording electronics to process the multiple inaudible signals to generate digital output samples, which are recorded sound data comprising non-linearities from frequency-shifted versions of the multiple inaudible signals to within an audible frequency range. A processing device is to detect, within the recorded sound data, at least a portion of the non-linearities, e.g., via: comparison of the recorded sound data with expected patterns from an audible audio signal generated by human voice; and detection of non-linear variations within the recorded sound data as compared to the expected patterns. In response to the detection, the processing device is further to suppress an action programmed for response to a voice command corresponding to the recorded sound data.

Causing a voice enabled device to defend against inaudible signal attacks

An audio transmitter includes a first ultrasonic speaker associated with a first channel; a second ultrasonic speaker co-located with the first ultrasonic speaker and associated with a second channel; and a waveform generator to: frequency modulate a first inaudible signal at a first ultrasonic frequency, to generate a modulated inaudible signal; drive, over the first channel, the first ultrasonic speaker with the modulated inaudible signal; and drive, over the second channel, the second ultrasonic speaker with a second inaudible signal at a second ultrasonic frequency so that a combination of the modulated inaudible signal and the second inaudible signal arrive at a microphone system. The second ultrasonic frequency is selected to frequency shift the modulated inaudible signal, upon demodulation by hardware of the microphone system, causing non-linearities of the hardware to translate the first ultrasonic frequency to below a low-pass filter cutoff frequency that is recordable by the microphone system.

Growing community of inventors

Champaign, IL, United States of America

Haitham Al Hassanieh