Granted Patent B2

US 12,257,949 · App. 17/648,928 · Granted Mar 25, 2025

Inventors: Karina Odinaev (Tel Aviv, IL); Igal Raichelgauz (Tel Aviv, IL)

Assignee: AUTOBRAINS TECHNOLOGIES LTD

B60Q9/00G06F3/165G10K15/04H04R1/08H04R3/00G08B3/10H04R2430/01H04R2499/13

Abstract

Claims (37)

1. A method for in-vehicle indication management, the method comprises:

obtaining, during a driving session of a vehicle, audio information about audio generated outside the vehicle;

determining, at a processor that comprises an integrated circuit, by processing the audio information, whether the audio generated outside the vehicle comprises driving affecting audio;

estimating, by the processor based on the determining, whether the driving affecting audio is heard within the vehicle; and

generating, in accordance with the estimating, a driving relevant in-vehicle audio indicative of the driving affecting audio, the generating involves searching, by applying a machine learning process trained to locate the driving affecting audio, for a signature matching between a generated signature of the audio and a generated signature for the driving affecting audio,

such that the driving relevant in-vehicle audio is generated, based on the searching, only when the vehicle is driven in a human driving mode, and further in accordance with the estimating, in response to estimating that the driving affecting audio is not being heard within the vehicle.

2. The method according to claim 1 wherein the driving relevant in-vehicle audio is a reconstructed version of the driving affecting audio.

3. The method according to claim 1 wherein the obtaining of the audio information comprising sensing the audio information by at least one external audio sensor.

4. The method according to claim 1 comprising estimating at least one other sound to be generated within the vehicle, when the driving relevant in-vehicle audio is played in the vehicle, wherein the at least one other sound differ from the driving relevant in-vehicle audio.

5. The method according to claim 4 comprising generating the driving relevant in-vehicle audio based on the estimating of the at least one other sound.

6. The method according to claim 4 comprising generating the driving relevant in-vehicle audio to be heard by at least the driver at the presence of the at least one other sound.

7. The method according to claim 4 wherein the at least one other sound is generated by a media device that is located, during the driving session, within the vehicle.

8. The method according to claim 1 , further comprising:

generating a signature of the audio information by:

performing a first number of signature generating iterations using a convolutional neural network of fixed connectivity; and

performing additional signature generation iterations that involve deactivating, per each additional signature generation iteration, spanning elements that are irrelevant to the additional signature generation iteration.

9. The method according to claim 1 , wherein the generating of the driving relevant in-vehicle audio comprising setting at least a volume of the driving relevant in-vehicle audio based on the driving mode.

10. The method according to claim 1 wherein the determining of whether the audio generated outside the vehicle comprises driving affecting audio is executed by a machine learning process that was trained to detect driving affecting audio.

11. A non-transitory computer readable medium that stores instructions that once executed by a processor that comprises an integrated circuit, cause the processor to:

obtain, during a driving session of a vehicle, audio information about audio generated outside the vehicle;

determine, by processing the audio information, whether the audio generated outside the vehicle comprises driving affecting audio;

estimate, based on the determining, whether the driving affecting audio is heard within the vehicle; and

generate, in accordance with the estimating, a driving relevant in-vehicle audio indicative of the driving affecting audio, by searching using a machine learning process trained to locate the driving affecting audio, for a signature matching between a generated signature for the audio and a generated signature for the driving affecting audio, such that the driving relevant in-vehicle audio is generated only when the vehicle is driven in a human driving mode, and further upon estimating that the driving affecting audio is not being heard within the vehicle.

12. The non-transitory computer readable medium according to claim 11 , that further stores instructions that once executed by the processor, causes the processor to:

generating a signature of the audio information by:

performing a first number of signature generating iterations using a convolutional neural network of fixed connectivity; and

performing additional signature generation iterations that involve deactivating, per each additional signature generation iteration, spanning elements that are irrelevant to the additional signature generation iteration; and

searching, by applying a machine learning process trained to locate the driving affecting audio, for a matching signature of the driving affecting audio.

13. A computerized system for in-vehicle indication management, the computerized system comprises a processor that comprises an integrated circuit and is configured to:

obtain, during a driving session of a vehicle, audio information about audio generated outside the vehicle;

estimate whether the driving affecting audio is heard within the vehicle; and

generate a driving relevant in-vehicle audio indicative of the driving affecting audio, only when the vehicle is being driven in a human driving mode, and when estimating that the driving affecting audio is not being heard within the vehicle.

14. The computerized system according to claim 13 , wherein the processor is further configured to:

generate a signature of the audio information by:

perform a first number of signature generating iterations using a convolutional neural network of fixed connectivity; and

perform additional signature generation iterations that involve deactivating, per each additional signature generation iteration, spanning elements that are irrelevant to the additional signature generation iteration; and

search, by applying a machine learning process trained to locate the driving affecting audio, for a matching signature of the driving affecting audio.

Assignments (1)

Continuity (2)

References Cited (400)