ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE - FOR DIAGNOSIS OF BOVINE RESPIRATORY DISEASES at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
2. ARTIFICIAL INTELLIGENCE (AI) INFUSED COW
NECKLACE - FOR DIAGNOSIS OF BOVINE RESPIRATORY
DISEASES
Presenter: Chandrasekar Vuppalapati
Hanumayamma Innovations and Technologies, Inc.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018
| Chengdu, China
AI – COW Necklace
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Authors: CHANDRASEKAR VUPPALAPATI, RAJASEKAR VUPPALAPATI, SHARAT KEDARI, ANITHA ILAPAKURTI, ARCHANA RAMALINGAM, JAYA SHANKAR
VUPPALAPATI, SANTOSH KEDARI
3. AI – COW Necklace
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
4. AI – COW Necklace
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
5. AI – COW Necklace
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
7. Who Are We
Ø Dairy Analytics is a wearable technology and analytics platform for Dairy Cattle.
Our technology gives Dairy Farmers data on the cattle's vital signs which is important
to a cows health and milk productivity.
ØDairy Analytics provides heat-detection, activity, and productivity insights.
ØDairy Analytics provides cattle activity based recommendations that include Milk
Fever, Ketosis, sick vs. healthy cattle detection and Lameness
ØDairy Analytics provides Forecasting modeling that helps Dairy management to
predict future operational costs and milk productivity.
December 25, 2016 7
Company
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
8. Why Dairy Analytics
ØWall Street Journal states “Data Software that monitors the cattle can reduce medication
costs by about 15% per animal and save more sick cattle from death.” By Jacob Bunge, Sept. 23,
2016
Ø Heat stress (HS) causes cows to produce less milk with the same nutritional input, which effectively
increases farmers’ production costs.
Ø The economic toll due to higher temperature, heat stress, is $1 billion annual problem. Not only in
the United States, but also around the globe, heat stress causes an adverse impact on the Dairy
productivity. [USDA]
Ø The U.S. Department of Agriculture estimates nearly $2.4 billion a year in losses from animal
illnesses that lead to death. This can be prevented by electronically checking on cattle’s vital signs.
8
Company
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
9. History
Ø Idea originated in February 2015
Ø Spoke to local and international dairy farmers [ USA, India, New Zealand] and test the idea of
Dairy Analytics to get feedback
Ø Traveled to cities in USA - Chicago, Gilroy and agriculture states in India - Punjab, Haryana, and
Telangana to show the sensor, screens and talk about Dairy Analytics with more Farmers
Ø Hired Engineers to start creating the Dairy Analytics Platform in April 2015
Ø Started first prototype of Dairy Internet of Things (IoT) Sensor in March 2015.
Ø Completed Development of the sensor and field tested in Gilroy, August 2016.
Ø Demoed the product in IEEE San Francisco, ADSA Chicago, and PDFA & Fatehgarh Sahib
Exposition, Punjab, India
Ø Exploring with prospective manufacturing vendors to scale up the unit production of the
Sensors – December 2016.
9
Dairy Analytics
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
10. Ø Dairy IoT Sensor – built to withstand high
temperatures and Humidity conditions
Ø Edge Analytics performed to relay real-time
notifications
Ø Water proof Casing & LED indicator for
Bluetooth Connectivity
Ø Ship setting configuration ensures over two
years Battery life
Ø Real-time capture of: Cattle body heat, Cattle
body humidity, Ambient Temperature,
Ambient Humidity
Ø Captures Cattle Motion Activity for analyzing
health patterns
10
Product
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
11. Product
11
Heat
Monitoring
Monitors activity related to
heats and identify optimal
windows for artificial
insemination
Dairy Analytics
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
12. Product
12
Health
Monitoring
Monitors behaviors such as
rumination, feeding, head
position and restlessness and
identify disease, lameness and
other health indicators.
Dairy Analytics
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
13. Presented at Intelligent Human Systems Integration: Integrating People and Intelligent Systems (iHSI 2018) on January 8, 2018 - 15:00 - 17:00, Dubai, UAE
13
Trademark Pending...
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
14. 14
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
15. Class 10: Wearable veterinary sensor for use in capturing
cow’s vital signs, providing data to the farmers to
measure the cow’s milk productivity, and improving its
overall health.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
16. 16
Demo
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
18. Ø A method for storing real-time dairy
sensor data streams in mobile devices.
Ø A method for transforming commercial
air fragrance dispenser into an
intelligent air fragrance dispenser
December 25, 2016 18
Engineering Architecture
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
19. Bovine respiratory disease (BRD) remains a significant cost
to both the beef and dairy industries. In the United States,
an estimated 640 million dollars is lost annually due to BRD.
.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
20. BRD, depending on the organism(s) involved, can cause
death within 24 to 36 hours of symptoms appearing, or the
infection can become chronic, not causing death but
instead producing widespread, permanent lung damage,
thus resulting major economic losses to the Dairy Industry .
.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
21. Given rapid progression and potential long-term & lethal
impact of the BRD, electronic health monitoring system,
albeit machine learning enabled Internet of Things (IoT)
powered wearable Dairy IoT Cow Necklace™, with prognosis
capabilities would play a pivotal role in preventing and / or
containing the disease.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
22. Electronically checking and real-time analyzing on cattle’s
heart rate, respiratory rate, digestion, temperature, cough
signatures and other vitals are vast advances that can cut
down on what the U.S. Department of Agriculture estimates
is nearly $2.4 billion a year in losses from animal illnesses
that lead to death.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
23. Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
24. Edge Analytics
The advantage of analyzing the data closer to the source
enables Edge Analytics not only provide rapid response but
also aids the detection of device health markers, data
anomalies and abnormalities so as to predict device
operational and/or health prognostics and thus potentially
improve the overall performance and life of the device.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
25. •Small Footprint Edge
o BLE Chip
o Accelerometer
o Time Sensor
o Proto Terminal Block
o EEPROM
o Microcontroller
o Power Supply
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
27. Ø Pattern Detection
Ø Historical Analysis
27
Machine Learning
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
28. Ø Sliding Window
Ø Edge Analysis
28
Machine Learning
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
29. Ø Pattern Detection
Ø Historical Analysis
29
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Machine Learning
30. Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
31. When an input is presented , the first layer computes
distances from the input vector to the training input vectors
and produces a vector whose elements indicate how close
the input is to a training input.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
32. The second layer sums these contributions for each class of
inputs to produce a vector of probabilities as its net output.
Finally, a competed transfer function on the output of the
second layer picks the maximum of these probabilities and
produces a 1 for that class and a 0 for the other classes.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
33. Signal à Pre-emphasis à Hamming Window à Fast Fourier Transform à Log à
cosine à Mel-frequency Cepstral coefficients à MFCC
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
34. Original Waveform:
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
35. Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
36. The mel-frequency cepstrum (MFCC) is the representation
of short-term power spectrum of a sound that is derived
from a linear cosine transform of a log power spectrum on a
nonlinear of a mel scale of frequency.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
37. Please note mel scale is a
scale of pitches with
reference point set to 1000
Hz tone, 40 dB above
listener’s threshold, with a
pitch of 1000 mels.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
38. Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
MFCC
39. Signal à Pre-emphasis à Hamming Window à Fast Fourier Transform à Log à
cosine à Mel-frequency Cepstral coefficients à MFCC
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
40. The first layer of our neural network represents audio
classification vectors that include: amplitude (Cn), end
point time interval (ψ), audio sensitivity threshold (ω),
noise levels (Ω) and Bluetooth signal interference
threshold (!).
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
41. Original Waveform:
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
42. End point detection (ψ) diff of red lines:
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
43. Filtered silence part: Ω = Org Sound wave – Filtered part
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
44. MFCC:
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Understanding Neural Networks for Dairy Sensor
45. Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
46. import glob
import os
import librosa
import librosa.display
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import tensorflow as tf
def extract_feature(row):
# To fetch the file name
audio_file = os.path.join(os.path.abspath( 'loadAudioFiles.wav')
try:
# Convert to numerical array
X, sampling_rate = librosa.load(audio_file, res_type = 'kaiser_fast')
# Extract 40 MFCC features from array
mfcc = np.mean(librosa.feature.mfcc(y=X, sr=sampling_rate, n_mfcc=40).T, axis=0)
except Exception as e:
print('Error while extracting MFCC feature:’)
return None, None
feature = mfcc
label = row.Class
return [feature, label]
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
47. Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Table Head
Table Column Head
Feature Label
0 [-82.1149459643,139.473175813, -42.4100851536...
Clinical
Case 1
1 [-15.7698946124, 124.144365738, -29.4644817551
Clinical
Case 2
2 [-237.933496285, 135.891856056, 39.25880357
Clinical
Case 3
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
48. The parameter in this layer is compared with Temperature (T), Humidity
(H), Activity (A), Medications (M), Digestive Health (D) and their
corresponding partial delta derivative vectors. That is, the vector holds the
last four-time interval change delta values.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
49. The parameter in this layer is compared with Temperature (T), Humidity
(H), Activity (A), Medications (M), Digestive Health (D) and their
corresponding partial delta derivative vectors. That is, the vector holds the
last four-time interval change delta values.
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
50. Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
51. Presented at Future of Information and Communication Conference (FICC) 2018, 5-6 April 2018 | Singapore
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
Thank You.
52. Management of Hanumayamma Innovations and Technologies, Inc., and
Hanumayamma Innovations and Technologies Private Limited for providing
Dairy IoT Sensors and Data...
Source:
Unlocking the potential of the internet of things - http://www.mckinsey.com/business-functions/digital-mckinsey/our-insights/the-internet-of-things-the-value-of-digitizing-the-physical-world
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
Acknowledgement
53. Machine Learning
Ø Pattern Detection
Ø Historical Analysis
December 25, 2016 53
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
54. Machine Learning
Ø Pattern Detection
Ø Historical Analysis
December 25, 2016 54
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
ARTIFICIAL INTELLIGENCE (AI) INFUSED COW NECKLACE
55. Machine Learning
Ø Pattern Detection
Ø Historical Analysis
December 25, 2016 55
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
56. Machine Learning
Ø Pattern Detection
Ø Historical Analysis
December 25, 2016 56
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China
57. AI – COW Necklace
Presented at the International Conference on Machine Learning and Cybernetics (ICMLC), 15-18 July 2018 | Chengdu, China