Smartphone recordings can determine whether a person has been drinking, and Stanford University research found that the accuracy is as high as 98%

According to reports, a new study from Stanford University shows that smartphones can accurately identify whether a person is drunk using voice patterns, with a recognition rate of up to 98%

The study tested 18 participants (72% male, aged 21-62), and sensors analyzed their voice patterns to detect whether they were drunk. The research was published in the "Journal of Studies on Alcohol and Drugs".

Brian Suffoletto, associate professor of emergency medicine at Stanford University, said he was surprised by the accuracy of the results, adding that further research is needed to confirm the validity of the findings.

He said the findings could help reduce future road injuries and deaths caused by, for example, drunk driving, adding: "While we are not the first team to study changes in speech characteristics during drunkenness, I firmly believe that our accuracy stems from our application of cutting-edge technologies in signal processing, acoustic analysis and machine learning."

This site learned from the research report that in this study , participants are given a dose of alcohol based on their body weight and must drink it within an hour. They were then put through a series of tongue twister tests. Participants had to repeat these out loud every hour for up to seven hours, while their voices were recorded by their smartphones.

Before drinking, the researchers also tested the participants’ breath alcohol content and recorded their tongue twisters. Breath alcohol content will then be tested every 30 minutes for 7 hours.

The researchers then used software to analyze the speaker's voice, looking at parameters such as frequency and pitch at one-second intervals, and built a support vector machine model to detect intoxication (defined as breathing alcohol Concentration > 0.08%), comparing the baseline speech spectrum characteristics with each subsequent time point and checking the accuracy of the 95% confidence interval (CI), resulted in a prediction accuracy of 98%.

Professor Suffoletto said a combination of steps and behavior such as texting could be used to determine a person's level of intoxication. He also added that the research results can be used to determine whether a person is drunk through mobile phone recordings, thereby providing timely intervention.

The above is the detailed content of Smartphone recordings can determine whether a person has been drinking, and Stanford University research found that the accuracy is as high as 98%. For more information, please follow other related articles on the PHP Chinese website!