Body camera footage storage creates new challenge for police agencies

STANFORD, Calif. (KGO) -- Body cameras are quickly becoming a game changer in the police world, but all the video they record creates a new challenge that some agencies are struggling to solve.

Officials from Oakland police and the White House spent the day meeting with experts.

Police body cameras capture violent confrontations like one where a woman who was eight months pregnant was wrestled to the ground.

They also captured an officer in Ohio who chose not to shoot, even after the suspect knocked him down.

In fact, that may be the untold story, says Oakland's police chief. "As we've grown our body worn camera program, we've seen very significant reduction in uses of force, we've seen significant reductions in complaints as well," Police Chief Sean Whent said.

The cameras hold police and citizens accountable. "This technology has deployed faster than any technology I've ever seen in my 45 plus years involved in policing," Major Cities Chiefs Police Assn.'s Darrel Stephens said.

But some large departments like San Francisco are struggling to implement the cameras.

"Future storage of the video files, the data, if you will, from the cameras. That could be very costly," SFPD Sgt. Michael Andraychak said.

"The Oakland Police Department currently is recording about 7-8 terabytes of data a month now," Whent said.

It all adds up to a pile of hard drives about like this created every single month, but more challenging than the space it takes up are the hundreds of hour it would take to watch all that footage.

"Where do we put it? How do we access it? How do we index it?" asked White House Chief Data Scientist D.J. Patil.

That's where Stanford comes in. "Ultimately, we'd like to be able to build some kind of speech recognition system," Stanford social psychologist Prof. Jennifer Eberhardt said.

Researchers have manually gone through thousands of traffic stops, logging key words and tone of voice.

"All of those things can give you a cue to whether an interaction is de-escalating or escalating," Eberhardt said.

A learning algorithm could one day scan videos automatically to quickly tell a department how well its officers are keeping the pace.

"Give training to an officer and then look at a sample after the training and you can measure the effectiveness of your training," Whent said.

The study's first findings could come out in March.