Abstract: A method includes receiving, as input to a speech recognition model, audio data corresponding to a spoken utterance. The method also includes performing, using the speech recognition model, speech recognition on the audio data by, at each of a plurality of time steps, encoding, using an audio encoder, the audio data corresponding to the spoken utterance into a corresponding audio encoding, and decoding, using a speech recognition joint network, the corresponding audio encoding into a probability distribution over possible output labels. At each of the plurality of time steps, the method also includes determining, using an intended query (IQ) joint network configured to receive a label history representation associated with a sequence of non-blank symbols output by a final softmax layer, an intended query decision indicating whether or not the spoken utterance includes a query intended for a digital assistant.

Abstract: Piloted or autonomous rotorcraft includes a rotor safety system. The rotor safety system comprises a lidar scanner toward a rotor of the rotorcraft, e.g., the tail rotor, that scans the 3D space in the vicinity of the rotor. Objects in the vicinity of the rotor are detected from the lidar point data. In a piloted rotorcraft, predictive warnings can be provided to the helicopter's flight crew when a detected object presents a hazard to the rotor of the rotorcraft.

Abstract: An apparatus comprises a time-of-flight ranging sensor that scans in two or more directions relative to the apparatus over a series of scanning cycles. A processor computes, and to communicates to the time-of-flight ranging sensor, a pulse repetition rate (PRR) for the time-of-flight ranging sensor for each of the two or more directions based on information about surrounding terrain of the apparatus and a sensor pointing schedule for the time-of-flight ranging sensor that indicates directions that the time-of-flight ranging sensor is scheduled to point at time during the scanning cycles. In addition or lieu of computing the PRR, the processor(s) matches returns from pulses of the time-of-flight ranging sensor to the pulses probabilistically based on a current map of the vehicle's surroundings and scan coherence analysis for shapes in the returns. The current map can then be updated based on the matched returns for the next iteration.

Abstract: A robotic sensor measures air quality characteristics as experienced by a toddler, such as a child of six to twelve months in age. The robot includes an air quality sensor, a terrain drive train, a sensor drive train and a control circuit that controls the terrain drive train and the sensor drive train. The control circuit directs the terrain drive train to traverse an area at a speed and a start and stop rate consistent with that of a child and directs the sensor drive train to control the monitoring height at which the air quality sensor measures the air quality characteristic in a manner consistent with that of the child.