Acoustic detection of glass breakage events
Systems and methods of using machine learning for identifying glass breakage events can include (a) accessing or obtaining the machine learning mode, (b) obtaining audio data, (c) determining, by processing the audio data using the machine learning mode, whether a glass breakage event has occurred, and (d) when determined that the glass breakage event has occurred, indicating the glass breakage event has occurred. Methods and instructions for training a machine learning model for identifying a glass breakage event can include (a) obtaining training audio data, (b) classifying the training audio data into a plurality of subsets, and (c) generating the machine learning model for identifying glass breakage events using the classified subsets.
1 . A computer-implemented method of using machine learning for
identifying glass breakage events, comprising:
accessing or obtaining, by one or more processors, a machine learning model trained using training audio data that represent (i) training ambient sounds including sounds caused by operation of machinery, (ii) training glass sounds including sounds caused by a first glass surface contacting a second surface which is either a second glass surface or a non-glass surface, wherein the first glass surface does not break or crack as a result of contacting the second surface, and (iii) training glass breakage sounds including sounds caused by either glass cracking or glass breaking;
obtaining, by the one or more processors, audio data over a period of interest;
determining, by the one or more processors processing the audio data using the machine learning model, whether a glass breakage event has occurred during the period of interest; and
when determined that the glass breakage event has occurred:
indicating, by the one or more processors, that the glass breakage event has occurred, and
automatically controlling, by the one or more processors, operation of the machinery.
2 . The computer-implemented method of claim 1 , wherein:
the training audio data and the audio data each include one or more spectrograms; and
obtaining the audio data over the period of interest includes generating the spectrogram of the audio data from raw audio data using a Fourier Transform.
3 . The computer-implemented method of claim 1 , wherein the machinery implements a biomanufacturing process and the training glass sounds and the training glass breakage sounds are produced by one or more containers of one or more drug products.
4 . The computer-implemented method of claim 1 , wherein automatically controlling operation of the machinery includes:
automatically causing, by the one or more processors, the machinery to stop operation.
5 . The computer-implemented method of claim 1 , further comprising:
iteratively adjusting, by the one or more processors, gain of amplification applied to one or both of training audio signals or audio signals until a performance threshold is satisfied, wherein the training audio signals and the audio signals respectively correspond to the training audio data and the audio data.
6 . The computer-implemented method of claim 1 , wherein
(a) the machine learning model is a convolutional neural network, and/or
(b) the machine learning model is trained using a supervised learning technique, and the training audio data include (i) ambient sound labels, (ii) glass sound labels, and (iii) breakage sound labels.
7 . A computer system of using machine learning for identifying glass breakage events, comprising:
one or more processors;
a program memory coupled to the one or more processors and storing executable instructions that, when executed by the one or more processors, cause the computer system to:
access or obtain a machine learning model trained using training audio data that represent, (i) training ambient sounds including sounds caused by operation of machinery (ii) training glass sounds including sounds caused by a first glass surface contacting a second surface which is either a second glass surface or a non-glass surface, wherein the first glass surface does not break or crack as a result of contacting the second surface, and (iii) training glass breakage sounds including sounds caused by either glass cracking or glass breaking,
obtain audio data over a period of interest,
determine, by processing the audio data using the machine learning model, whether a glass breakage event has occurred during the period of interest, and
when determined that the glass breakage event has occurred:
indicate that the glass breakage event has occurred, and
automatically control operation of the machinery.
8 . The computer system of claim 7 , wherein:
the training audio data and the audio data each include one or more spectrograms; and
obtaining the audio data over the period of interest includes generating the spectrogram of the audio data from raw audio data using a Fourier Transform.
9 . The computer system of claim 7 , wherein the machinery implements a biomanufacturing process and the training glass sounds and the training glass breakage sounds are produced by one or more containers of one or more drug products.
10 . The computer system of claim 7 , wherein to automatically control operation of the machinery, the instructions, when executed by the one or more processors, further cause the computer system to:
automatically cause the machinery to stop operation.
11 . The computer system of claim 7 , wherein the machine learning model is a convolutional neural network.
12 . The computer system of claim 7 , wherein:
the machine learning model is trained using a supervised learning technique, and
the training audio data include, (i) ambient sound labels, (ii) glass sound labels, and (iii) breakage sound labels.
13 . The computer system of claim 7 , wherein, the instructions, when executed by the one or more processors, further cause the computer system to:
iteratively adjust gain of amplification applied to one or both of training audio signals or audio signals until a performance threshold is satisfied, wherein the training audio signals and the audio signals respectively correspond to the training audio data and the audio data.
14 . A computer-implemented method for training a machine learning model for identifying glass breakage events, comprising:
obtaining, by one or more processors, training audio data;
classifying, by the one or more processors, the training audio data into a plurality of subsets each corresponding to different actual outcome data, the subsets including (i) at least one subset representing training ambient sounds including sounds caused by operation of machinery, (ii) at least one subset representing training glass sounds including sounds caused by a first glass surface contacting a second surface which is either a second glass surface or a non-glass surface, wherein the first glass surface does not break or crack as a result of contacting the second surface, and (iii) at least one subset representing training glass breakage sounds including sounds caused by either glass cracking or glass breaking; and
generating, by the one or more processors, the machine learning model for identifying glass breakage events using the classified subsets of the training audio data.
15 . The computer-implemented method of claim 14 , wherein the training audio data include one or more spectrograms.
16 . The computer-implemented method of claim 14 , wherein the machinery implements a biomanufacturing process and the training glass sounds and the training glass breakage sounds are produced by one or more containers of one or more drug products.
17 . The computer-implemented method of claim 14 , further comprising:
iteratively adjusting, by the one or more processors, gain of amplification applied to training audio signals until a performance threshold is satisfied, wherein the training audio signals correspond to the training audio data.
18 . The computer-implemented method of claim 14 , wherein the machine learning model is a convolutional neural network.