Prioritizing Emergency Calls: How Can Dispatchers Use Voice Stress Analysis?
The Computer Voice Stress Analyzer (CVSA®) has shown great value as a tool for law enforcement professionals to detect deception. It is commonly used to help solve criminal cases, screen job candidates, conduct terrorism-related military investigations, and evaluate immigration cases. However, voice stress analysis has also been applied to other pressing problems. For instance, a study conducted in 2012 by Polish researchers Grazyna Demenko and Magdalena Jastrzebska found that analyzing speech for signs of stress could be a means of prioritizing emergency calls.
These researchers examined key relationships between voice pitch and stressful situations. Using Multidimensional Voice Program (MDVP) software, a program that allowed them to analyze voices for a wide range of acoustical features, the researchers established that stress levels directly affect speech. Contemporary CVSA technology uses a specialized voice program to detect vocal stress indicative of deception, giving law enforcement officers a valuable tool to screen job candidates, assess witnesses and suspects, and gather information. In this study, researchers found that voice pitch could potentially help emergency dispatchers make better choices in applying response resources.
The Study’s Background
Emergency dispatchers receive many different calls with varying levels of urgency. For instance, when someone calls to report that their house was broken into while they were at work, it is less urgent than when someone calls to report their friend is having a heart attack. Demenko and Jastrzebska hypothesized that if callers’ voices could be analyzed for signs of stress, emergency operators could make better life-saving choices about which calls to prioritize.
Research conducted prior to the study indicated there were certain acoustic features of the voice that could be associated with particular emotional states. For instance, emotions like fear, anger, and joy could trigger the sympathetic nervous system, resulting in loud, fast, and/or highly enunciated speech. Conversely, emotions like boredom and sadness could trigger the parasympathetic nervous system, leading to slow, low-pitched speech. It had also been demonstrated that stress can lead to changes in the fundamental voice frequency, which is undetectable by ear but can be measured with CVSA technology.
Analyzing Emergency Calls
To conduct their experiment, the researchers combed through a database of over 8,000 calls to emergency dispatchers for a wide range of reasons, including crime notifications and requests for police intervention. Each call was annotated based on the background acoustics, the type of the dialogue, and key features of the caller’s speech like rate, volume, intonation, context, duration, and general emotion. In order to facilitate comparisons of voice pitch between stressful and non-stressful situations, the researchers limited the study sample to include only calls where the caller had demonstrated more than one level of stress. To make this determination, they ranked arousal levels on a numerical scale and only chose samples where there was at least one dialogue at level 0 and one dialogue that was at 0.5 or above.
The aspects of pitch the researchers measured were pitch position and pitch compass. Pitch position describes the observed pitch of the caller’s voice at any given moment, as compared to the minimum pitch (Fmin) over the course of the dialogue. Pitch compass refers to the overall pitch range (the difference between Fmin and the maximum pitch Fmax). Fmin and Fmax were automatically measured using voice software. Demenko and Jastrzebska found that pitch position and pitch compass varied based on the stressfulness of a situation, and they described several relevant test cases to illustrate their findings:
- One caller called twice: once to report a burglary (a call made under stress) and again to call of police intervention (a non-stressful situation). While the pitch position did not change over the course of each individual dialogue, the pitch position in the high-stress dialogue was 40 Hz higher than in the low-stress example.
- In one example, a masked person had entered the caller’s apartment—a direct threat. As the intruder moved closer and the threat became more immediate, the caller’s pitch position shifted up. When the dispatcher instructed the person to calm down, however, the pitch position returned to a lower level. Although the intruder was still present (the threat had not, in fact, subsided), the dispatcher’s directions likely helped reduce the caller’s stress levels.
- The researchers also evaluated calls where the pitch position and the pitch compass changed between utterances. They found this was typically associated with anger and mixed emotions.
- Similarly, the study found that when a caller went from a neutral state to an angry state, the pitch position may remain the same, but the pitch compass was subject to change.
Predicting Stress Levels in Emergency Callers
Based on the data they collected, as well as previous findings on factors that affect voice stress, the authors used Linear Discriminant Analysis (LDA), a generalized statistical classification method, to divide a group of voice samples into four categories: male – stress, female – stress, male – neutral/mild irritation, and female – neutral/mild irritation. Ultimately, their program had an accuracy rate of 80 to 84 percent. Thus, they concluded that voice analysis is a valuable tool for prioritizing emergency calls.
Demenko and Jastrzebska’s scientific study serves as yet another assurance of the usefulness and accuracy of voice stress analysis technology. Law enforcement officials who have used CVSA in the field already know the power of this technology to detect deception and advance investigations, but these findings prove that voice stress analysis has practical applications both within and outside the field of law enforcement.