Using six seconds of speech, software developed by a company called Tenvos AI claims to tell whether workers are too sleep-deprived to safely operate heavy machinery or drive a large truck on the highway.

The creators of the technology hope it will cut down on workplace injuries, 13% of which are linked to fatigue, according to Tenvos AI. It also checks for alcohol and cannabis use, but in an interview Thursday, founder and CEO Rima Seiilova-Olson primarily focused on its implications for assessing exhaustion.

“There is really no objective measure that can quantifiably say, ‘OK, you’re tired,'” she said in a video call. “We can approximate it. Even I myself, it’s really hard for me to judge what my fatigue level is, so creating an objectively quantifiable metric around fatigue is the gap in the market, especially for industries where safety is paramount.”

Being awake for 17 hours is similar to having a blood alcohol level of 0.05%, according to guidance on the U.S. Centers for Disease Control website. That rises to the equivalent of 0.1% once someone is awake for 24 hours, which is above the legal limit of 0.08% in Tennessee.

(READ MORE: Autonomous electric vehicle charging start-up plans move to Chattanooga)

While there are biological means of measuring drug and alcohol impairment, Seiilova-Olson said fatigue isn’t as easy to quantify. Multiple factors can determine how much sleep someone needs.

“If someone’s drunk, we don’t let them work, especially at a petrochemical plant or God forbid drive on the highway, right?” Seiilova-Olson said. “The level of impairment induced by alcohol intoxication at 0.08% is almost the same as either 18 or 20 hours of wakefulness, but we let sleepy people go into work because we just can’t tell.”

The start-up recently participated in The Company Lab’s sustainable mobility accelerator, a 12-week program in Chattanooga that provides fledgling businesses with access to corporate sponsors, a network of mentors and a $20,000 investment with the potential for an extra $100,000. The program generally picks up more than 100 applicants from around the globe.

Tasia Malakasis, CEO of The Company Lab, said the program only accepts startups if one of the organization’s corporate, research or municipal partners indicate the business has a solution that addresses one of their critical problems. The Company Lab heard resoundingly from its partners that detecting impairment is a real need, she said.

“For logistics, transportation and manufacturing, I don’t know that anyone has solved the problem that exists with keeping the workplace safe and keeping workers safe,” Malakasis said in a phone call. “When you think about what kind of damage can a truck driver do — it’s not just to them and to that truck — that’s a tremendous liability if you’re running into anything else. She’s solving a really critical problem.”

Xinglei Liu, a machine learning engineer with Tenvos AI, said producing speech requires precise motor skills that involve controlling hundreds of little muscles. If someone is exhausted, their pronunciation of distinct phonemes — the smallest unit of speech in the English language — will be different.

“Previous speech scientists have found a lot of muscle changes related to fatigue,” Liu said in a video call. “When people are impaired, their motor control of their lips and tongue and jaw will be different.”

Tenvos AI first gets a personalized baseline from workers so they can establish how that individual normally sounds.

The specific application of the technology depends on the setting.

Truck drivers could have the technology implemented as an application on a computer tablet in their vehicle. Employees at a stationary manufacturing site may walk up to a kiosk, punch in their employee number and read a couple sentences that are then analyzed by the software. For example, the phrases could include, “Together we achieve more and reach new heights. The source of the huge river is the clear spring.”

“These sentences are not random,” Seiilova-Olson said. “They have every single phoneme in the English language in them, so they’re kind of a condensed version of the English language.”

If fatigue is identified in the employee’s voice, an email is sent to the worker and manager, who can then have a conversation about next steps, Seiilova-Olson said. Nine out 10 times, workers are dealing with external circumstances like managing elderly parents, caring for a newborn or dealing with a time change that lost them an hour of rest, she said.

In those cases, someone normally tasked with driving a forklift, for example, could switch roles with another coworker for the day, Seiilova-Olson said. An employee in a management position could be paired with someone who can provide a second opinion on decisions. Some petrochemical and mining companies also have a nap room.

“Depending on the job responsibilities, there are different ways of either redesigning the workflow or simply administering countermeasures,” Seiilova-Olson said.

Seiilova-Olson said the sustainable mobility accelerator gave Tenvos AI a greater appreciation of how trucking and logistics could fit into its customer base, which also includes construction, manufacturing and oil and gas companies. Every truck is now like a smartphone, Seiilova-Olson said, outfitted with cameras, electronic logging devices and GPS.

“They’re going to be equipped with so much intelligence that will improve safety on our highways, and the sustainable mobility accelerator helped us realize how mature the field is,” she said.

(READ MORE: CEO of Chattanooga-based Covenant Logistics optimistic about Trump economy)

The company is now working on picking up customers in and around Hamilton County.

“I’m not sure we can disclose the names of the companies because we haven’t negotiated that specific clause in the contract, but … we have at least four use cases that we’re actively pursuing for paid pilots,” Seiilova-Olson said. “Two of them should be kicking off relatively soon.”

Addressing privacy concerns that can crop up with the technology, Seiilova-Olson said the company works to transparently explain to customers how their data is used.

“We, for instance, entirely strip out the semantic meaning of the words that are being said,” she said. “We really don’t care about what people are saying. All we care about is how.”

Contact business reporter David Floyd at dfloyd@timesfreepress.com or 423-757-6249.