Vehicle with external audio speaker and microphone

A method for audio events detection in a vehicle, the method comprising:

• receiving audio data, the audio data sensed by a microphone array installed on the vehicle;
• analyzing the audio data using a machine learning technique to recognize a sound event, wherein analyzing the data includes to determining that the sound event represents a verbal call to an operator of the vehicle by a person outside of the vehicle;
• in response to a successful recognition of the sound event from the audio data:
• determining, based on analyzing the audio data, whether the verbal call relates to a safety condition;
• in response to determining that the verbal call relates to the safety condition, causing a vehicle control device to change a trajectory of the vehicle;
• and
• causing the vehicle to emit a communication to the person outside of the vehicle;
• in response to an unsuccessful recognition of the sound event from the audio data in the audio data, identify or label the audio data as containing the sound event from a user input, and wherein, in response to the audio data being labeled as containing the sound event, the machine learning technique is updated with the labeled audio data to identify a similar sound event in a subsequent classification;
• and
• transmitting a message to a vehicle control system, the message based on the sound event.

At least one machine-readable medium including instructions for audio events detection in a vehicle, the instructions when executed by a machine, cause the machine to perform operations comprising:

• receiving audio data, the audio data sensed by a microphone array installed on the vehicle;
• analyzing the audio data using a machine learning technique to recognize a sound event, wherein analyzing the data includes determining that the sound event represents a verbal call to an operator of the vehicle by a person outside of the vehicle;
• in response to a successful recognition of the sound event from audio data:
• determining, based on analyzing the audio data, whether the verbal call relates to a safety condition;
• in response to determining that the verbal call relates to the safety condition, causing a vehicle control device to change a trajectory of the vehicle;
• and
• cause the vehicle to emit a communication to the person outside of the vehicle;
• in response to an unsuccessful recognition of the sound event from audio data, identify or label the audio data as containing the sound event from a user input, and wherein, in response to the audio data being labeled as containing the sound event, the machine learning technique is updated with the labeled audio data to identify a similar sound event in a subsequent classification;
• and
• transmitting a message to a vehicle control system, the message based on the sound event.