Presented By O'Reilly and Cloudera
Make Data Work
March 13–14, 2017: Training
March 14–16, 2017: Tutorials & Conference
San Jose, CA

Real-time analysis of behavior of law enforcement encounters using big data analytics and deep learning multimodal emotion-recognition models

Nixon Patel (Kovid Group)
3:30pm4:00pm Tuesday, March 14, 2017
DCS, Strata Business Summit
Location: LL20 A Level: Intermediate
Secondary topics:  Deep learning
Average rating: *....
(1.40, 5 ratings)

Sometimes, encounters between law enforcement officers and the public escalate, resulting in use of force by the officers. A number of jurisdictions have acknowledged the importance of de-escalating interactions when they appear to be getting out of control. A great deal of emphasis is placed on training the officers to handle stressful encounters in a calm manner. In a few cases, patrol officers are also doubled up, allowing the second officer to step in and calm things down.

While some situations represent genuine threats to an officer (where the officers’ responses are justified), some situations can be attributed to the officer’s emotional state. If the emotional state of officers could be monitored, it might be possible to prevent these incidents.

Nixon Patel offers an overview of deep learning and AI models that capture and analyze the emotional state of law enforcement officers over a period of time, based on anonymized voice recordings of law enforcement encounters gathered over a period of time and conducted with the authorization of appropriate agencies, enabling an officer’s supervisors to intervene if a given officer is subject to repeated emotional stress.

The system extracts emotional cues from a combination of a subject’s facial expressions and voice. The system interprets these expressions and reads the emotional state of officers when they are in front of their mobile data terminal (MDT). In addition to the direct measurement of the emotional state of the officer (and the member of the public they are interacting with), certain keywords or phrases could also indicate an escalating level of stress in the interaction. Keyword detection is applied, with words and phrases selected using domain experts, and added to the input vector. The input to the models also included demographic information, the seniority of the officer, the time elapsed since the last training, and characteristics of the patrol location. Behavior models are updated with each encounter and the progression of the emotional state of the officers is estimated at each state.

This information is used in several ways:

  • Visualization on dashboard to the supervisors, who will be provided with alerts if the emotional state of one of their officers is deemed to have deteriorated
  • Alerts to the officers themselves, displayed on the MDT
  • Peer-to-peer support (Supervisors have the option to direct peers who happen to be nearby to intervene.)
  • Self-monitoring

Because the system is able to work in a proactive mode, violent encounters involving the unjustified use of force by the officers could be reduced. This, in turn, will help the police establish trust and confidence between the police and the communities.

Photo of Nixon Patel

Nixon Patel

Kovid Group

Nixon Patel is founder, CEO, and MD of analytics startup Kovid. Nixon is a visionary leader, an exemplary technocrat, and a successful, business-oriented entrepreneur with a proven track record for growing six businesses from startups to large, global technology companies with millions in annual sales—in industries ranging from big data technology, analytics, the cloud, the IoT, speech recognition, and machine learning to renewable energy, information technology, telecommunications, and pharmaceuticals—all over a short 26-year span. Previously, he was chief data scientist at Brillio, a sister company of Collabera, where he was instrumental in starting the big data analytics, cloud, and IoT practices and establishing centers of excellence and co-innovation labs. He is also an independent director in VivMed Labs and Tripborn. Nixon holds a BT with honors in chemical engineering from IIT Kharagpur, an MS in computer science from the New Jersey Institute of Technology, and a data science specialization from Johns Hopkins University. He is currently pursuing a second master’s degree in business and science in analytics from Rutgers University.