How Did EMG Experiments Lead to Contemporary Voice Stress Tests?
Over the course of about a decade, J. Gary Eden and Gideon F. Inbar—two researchers in the Department of Electrical Engineering Technology at the Israel Institute of Technology—conducted multiple studies in which they used electromyography (EMG) to study voice tremor. Their results indicated that voice tremor is correlated with stress, leading them to conclude it is possible to detect stress by analyzing voice recordings. This research served as verification for some of the initial voice stress analysis technologies being developed at the time. Overall, Inbar and Eden’s body of work serves as a scientific basis for the Computer Voice Stress Analyzer® (CVSA), which relies on the principles Inbar and Eden elucidated using EMG.
Understanding the Electromyography Experiments of Inbar and Eden
In EMG experiments, researchers record the electrical activity in a muscle or group of muscles. This activity stems from minute oscillations in muscles known as muscular tremor or physiological tremor. Inbar and Eden were particularly interested in correlating muscular tremor in the vocal tract with variations in voice frequency.
To set up an electromyography experiment, researchers attach electrodes to the skin adjacent to the muscles of interest. For their experiments, Inbar and Eden placed the electrodes along the neck and near the extrinsic muscles of the tongue. That way, when the subject spoke, the display could produce a visual depiction of the oscillations in the vocal tract. At the same time, Inbar and Eden also recorded the frequency of the subject’s voice so they could compare the two charts.
When taking their measurements, Inbar and Eden instructed their subjects to speak one of the five phonemes (vowels)—A, E, I, O, or U—and hold the sound for several seconds while attempting to maintain a constant pitch and general vocal tract configuration. This structure became the basis for the EMG experiments that would lead them to author three key studies that feature prominently in the history of voice stress analysis.
A Decade of Electromyography (EMG) Experiments
Inbar and Eden started studying voice tremor together in the mid-1970s, and their research continued into the mid-1980s. During that decade, the duo published three peer-reviewed papers:
- “Physiological Stress Evaluators: EMG Correlation with Voice Tremor” (1976)
- “Physiological Model Analysis of Involuntary Human-Voice Tremor” (1978)
- “Physiological Evidence for Central Modulation of Voice Tremor” (1983)
In their 1976 paper, Inbar and Eden reported on a three-part study in which their goal was to verify that voice frequency changes truly correlated with physiological tremor, as measured by an EMG. For this, they performed the EMG experiments described above, generating two types of signals: sound waves from the spoken vowels (voice frequency) and EMG waves recorded while the subjects were making the speech sounds (physiological tremor). From there, they developed a complex mathematical “filtering” method, which they used to verify that there was a statistically significant connection between voice frequency and physiological tremor.
In addition, Inbar and Eden explored the hypothesis that voice frequency changes were controlled by the central nervous system (CNS). This connection to the CNS would suggest that changes in the brain—including those that occur when a person is under stress—could directly impact voice frequencies. To test this hypothesis, they used EMG to stimulate vocal tract muscles with periodic electrical impulses (as a way to experimentally simulate stimulation by the CNS) while the subject was speaking. As expected, they found that the stimulation modulated voice frequency.
In their 1978 study, Inbar and Eden asked the next obvious question: How exactly was the involuntary CNS stimulation causing changes in voice frequency? And what physiological mechanisms were behind the relationship? By that point, early technologies to detect deception by measuring vocal stress had been developed, and Inbar and Eden hoped to use their study to either verify the validity of these technologies or discredit them with physiological evidence.
To learn more about CNS stimulation and voice frequency, Inbar and Eden used accepted mathematical models of the vocal tract and vocal cords and isolated the parameters most likely to influence involuntary changes in voice frequencies. Using these models, they were able to explain their previous experimental results by showing that the physiological effects most responsible for voice frequency changes were geometrical changes of the vocal tract and tension in the vocal cords. Both of these can be involuntarily induced by stress, so Inbar and Eden research supports the idea that voice stress analysis is indeed a reliable method for detecting stress associated with deception.
Inbar and Eden’s 1983 report reconfirmed and built on their previous findings regarding the CNS’ role in facilitating the relationship between physiological tremor and voice frequency. First, the researchers repeated their experiments from 1976 to verify that there is, indeed, a statistically significant correlation between voice frequency and physiological tremor. Then, they conducted EMG tests on two groups of patients with compromised CNS function: a group of patients with Parkinson’s disease and a group of patients with traumatic brain injuries. They found that the level of voice tremor—that is, the amount of muscle oscillation recorded when the patient spoke—was significantly higher than normal in patients with Parkinson’s and significantly lower than normal in patients with traumatic brain injuries. From this, they were able to conclude that physiological tremor is directly dependent on involuntary CNS activity.
Implications for Voice Stress Analysis Technology
Ultimately, Inbar and Eden’s studies demonstrated a significant correlation between physiological tremor and voice frequency. By using both mathematical models and clinical studies, they were also able to show that the level of physiological tremor is determined by involuntary CNS activity, including stress. Thus, these researchers concluded that voice stress analysis technologies—including CVSA—that use physiological tremor as a means of determining an individual’s stress levels were effective technological tools for identifying possible deception.
Law enforcement officials who have used CVSA in the field already know the technology’s value when interviewing suspects and witnesses, but it’s important to also understand the technology’s scientific background. After all, the rigorous research of Inbar and Eden has stood the test of time and still remains some of the most important evidence for the reliability of CVSA in law enforcement efforts to identify stress and detect deception.