Gunshot detection via classification using deep learning and fiber sensing technologies
A machine learning (ML)/artificial intelligence (AI) based distributed fiber optic sensing (DFOS) system and method providing detection and localization of gunshot events. In addition to the ML/AI DFOS, a signal processing pipeline that compresses an audible distributed acoustic sensing (DAS) waveform data into a small set of features that protects privacy of individuals while preserving the utility of acoustic events to detect the gunshot events and discriminate same from other events. A data-driven deep learning approach automatically predicts acoustic event types with higher accuracy that realized by prior art methods.
1 . A method for detecting impulsive events using distributed fiber optic sensing (DFOS), the method comprising:
providing the DFOS system including:
a length of optical sensing fiber;
an optical interrogator configured to generate optical pulses, introduce the generated optical pulses into the optical sensing fiber and receive Rayleigh reflected signals from the optical sensing fiber;
an artificial intelligence (AI) analyzer configured to analyze the Rayleigh reflected signals received by the optical interrogator and determine the existence and nature of the impulsive events;
operating the DFOS system and determining existence and nature of the impulsive events; and
distinguishing, from their determined nature, impulsive events requiring an intervention from impulsive events requiring no intervention.
2 . The method of claim 1 wherein the provided DFOS system length of optical sensing fiber includes both aerial and underground lengths of optical sensing fiber.
3 . The method of claim 2 wherein the DFOS system length of optical sensing fiber includes one or more fiber coils and one or more fiber-based acoustic enhancers.
4 . The method of claim 3 wherein the received Rayleigh reflected signals are processed into a time series data.
5 . The method of claim 4 wherein the time series data undergo a Short-time Fourier Transform (STFT) processing producing STFT data.
6 . The method of claim 5 where the STFT data is processed to extract Mel-frequency cepstral coefficients (MFCCs).
7 . The method of claim 6 wherein the MFCCs are presented to a convolutional neural network (CNN) for classification.
8 . The method of claim 7 wherein classifications produced by the CNN are used to determine an intervention.
9 . The method of claim 8 wherein the DFOS system comprises an additional length of optical sensing fiber, optically isolated from the length of optical sensing fiber.
10 . The method of claim 9 wherein the DFOS system includes an optical switch interposed between the optical interrogator, the length of optical fiber and the additional length of optical sensing fiber.
11 . A system for detecting impulsive events using distributed fiber optic sensing (DFOS), the system comprising:
the DFOS system including:
a length of optical sensing fiber;
an optical interrogator configured to generate optical pulses, introduce the generated optical pulses into the optical sensing fiber and receive Rayleigh reflected signals from the optical sensing fiber;
an artificial intelligence (AI) analyzer configured to analyze the Rayleigh reflected signals received by the optical interrogator and determine the existence and nature of the impulsive events and distinguish, from their determined nature, impulsive events requiring an intervention from impulsive events requiring no intervention.
12 . The system of claim 11 wherein the provided DFOS system length of optical sensing fiber includes both aerial and underground lengths of optical sensing fiber.
13 . The system of claim 12 wherein the DFOS system length of optical sensing fiber includes one or more fiber coils and one or more fiber-based acoustic enhancers.
14 . The system of claim 13 further configured to process the received Rayleigh reflected signals into a time series data, process the time series data via a Short-time Fourier Transform (STFT) processing producing STFT data, and processing the STFT data to extract Mel-frequency cepstral coefficients (MFCCs).
15 . The system of claim 14 further configured to present the MFCCs to a convolutional neural network (CNN) for classification.
16 . The system of claim 15 wherein classifications produced by the CNN are used to determine an intervention.
17 . The system of claim 16 further comprising an additional length of optical sensing fiber, optically isolated from the length of optical sensing fiber.
18 . The system of claim 17 further comprising an optical switch interposed between the optical interrogator and the length of optical fiber and the additional length of optical sensing fiber.