Sensor fusion for collision detection

A method for collision detection, comprising:

• receiving a first sensor data from a first sensor and a second sensor data from a second sensor, wherein a plurality of sensors includes the first sensor and the second sensor, wherein the first sensor includes an audio sensor, wherein the second sensor comprises one or more of the following:
• an accelerometer, a gyroscope, and/or a GPS, wherein the first sensor is different from the second sensor, wherein the first sensor data includes audio data, and wherein the first sensor data is different from the second sensor data;
• preprocessing, using a processor, the first sensor data using a set of preprocessing steps to create first preprocessed sensor data, wherein the set of preprocessing steps comprises filtering, normalization, and alignment;
• preprocessing the second sensor data using the set of preprocessing steps to create second preprocessed sensor data;
• generating, using a first model, a first neural network output based on the first preprocessed sensor data, wherein the first model implements a first neural network technique, wherein the first model is trained using model data tagged with a collision indication indicating whether a crash or not occurred;
• generating, using a second model, a second neural network output based on the second preprocessed sensor data, wherein the second model implements a second neural network technique, wherein the second model is trained using the model data tagged with a collision indication indicating whether a crash or not occurred;
• processing the first neural network output and the second neural network output using a compound model to create a collision score, and wherein the processing of the first neural network output and the second neural network output comprises:
• combining the first neural network output and the second neural network output as input to the compound model to obtain the collision score, wherein the combining of the first neural network output and the second neural network output includes performing an exclusive OR (XOR) of the first neural network output and the second neural network output;
• and
• providing the collision score;
• and
• performing an action based at least in part on the collision score, wherein the action comprises one or more of:
• initiating a 911 call, initiating driver contact for a client operator, providing a tow truck request, providing a backup truck for cargo transfer request, or providing an ambulance request.

A computer program product for collision detection, the computer program product being embodied in a non-transitory computer readable storage medium and comprising computer instructions for:

• receiving a first sensor data from a first sensor and a second sensor data from a second sensor, wherein a plurality of sensors includes the first sensor and the second sensor, wherein the first sensor is different from the second sensor, wherein the first sensor includes an audio sensor, wherein the second sensor comprises one or more of the following:
• an accelerometer, a gyroscope, and/or a GPS, wherein the first sensor data includes audio data, and wherein the first sensor data is different from the second sensor data;
• preprocessing the first sensor data using a set of preprocessing steps to create first preprocessed sensor data, wherein the set of preprocessing steps comprises filtering, normalization, and alignment;
• preprocessing the second sensor data using the set of preprocessing steps to create second preprocessed sensor data;
• generating, using a first model, a first neural network output based on the first preprocessed sensor data, wherein the first model implements a first neural network technique, wherein the first model is trained using model data tagged with a collision indication indicating whether a crash occurred or not;
• generating, using a second model, a second neural network output based on the second preprocessed sensor data, wherein the second model implements a second neural network technique, wherein the second model is trained using the model data tagged with a collision indication indicating whether a crash or not occurred;
• processing the first neural network output and the second neural network output using a compound model to create a collision score, and wherein the processing of the first neural network output and the second neural network output comprises:
• combining the first neural network output and the second neural network output as input to the compound model to obtain the collision score, wherein the combining of the first neural network output and the second neural network output includes performing an exclusive OR (XOR) of the first neural network output and the second neural network output;
• and
• providing the collision score;
• and
• performing an action based at least in part on the collision score, wherein the action comprises one or more of:
• initiating a 911 call, initiating driver contact for a client operator, providing a tow truck request, providing a backup truck for cargo transfer request, or providing an ambulance request.