The jitter and shimmer quality of one’s voice may be a greater indicator of marital success than the input provided by relationship experts, a recent study from the Univ. of Southern California found.
A team from the university’s Viterbi School of Engineering and the Univ. of Utah designed a computer algorithm that used speech-processing techniques to break down recordings from marriage counseling sessions into acoustic features. Impressively, the algorithm was capable of predicting whether a couple’s relationship would better or worsen with nearly 79% accuracy.
“The marital outcomes of a distressed couple undergoing counseling has been a subject of long standing interest in order to assess the quality and the effectiveness of marriage counseling sessions,” write the researchers in Proceedings of Interspeech. “An objective demonstration of the quality of couple therapy and the progress of the couple towards a better relationship can be helpful to the domain experts in evaluating and adjusting their therapy methods.”
The researchers used a couple therapy corpus stemming from a larger longitudinal study performed by the Univ. of California, Los Angeles and the Univ. of Washington. One-hundred thirty-four distressed couples, married on an average of 10 years, participated in the study. The audio-visual recorded and coded sessions took place before therapy started, after 26 weeks of therapy and after two years of therapy. On top of jitters and shimmers, intensity, pitch, warbles and other qualities were measured.
For comparison, the research team tested the algorithm against behavioral analyses completed by human experts, who coded by positive and negative qualities, such as “acceptance” or “blame.”
“Acoustic features capture more relevant information than the human-encoded behavioral constructs for predicting marital outcomes,” they write.
According to coauthor Brian Baucom, of the Univ. of Utah, efficient and reliable tools for measuring the minutia of a conversation are lacking. “These findings represent a major step forward in making objective measurement of behavior practical and feasible for couple therapists,” he said.
Next, the team plans on using the computational framework, developed by study lead Shrikanth Narayanan, to predict how effective treatments will be.