Speech analysis is the process of analyzing the voice and the speech patterns of people. Some approaches of speech analysis are focused on speech recognition - understanding the speech content. Other approaches are focused on speech prosody - understanding the speech “melody” or the pronunciation patterns, typically analysing speech aspects of intonation, rhythm, stress, emphasis.
Speech prosody is strongly related to the physiology of voice and speech production, involving the vocal cords, breathing, pitch, loudness, air flow patterns and other acoustic features. Prosody is used to express the speaker’s attitude, similar to facial expressions – serious, light, dramatic, warm, childish, ironic, urgent etc. Based on physiology and focused on attitude rather than content, speech prosody is generally language independent.
VoiceSense has taken this approach a couple of steps further.
Speech Profiling – A Unique Biometric Concept
VoiceSense introduces a new biometric concept – personal profiling through speech analysis. The concept is based on measuring typical speech patterns within a person’s natural interactions. These patterns reflect emotional responses and behavioral tendencies such as temperament, determination, sociability, openness, adaptability. When measured in a given interaction, these patterns reflect the emotional sentiment and the current state of mind of a person. When measured consistently over time, they reflect personality and well-being.
By extracting the generic speech prosody features, we provide not only a language independent analysis but also a speaker independent tool (no need for training or calibrating to the specific voice) for understanding people’s emotions and attitudes (state of mind).
And by adding the characteristic prosodic speech patterns of an individual, we are able to provide personality profiling.
Highlights of Our Approach
Typical methods for sentiment and personality measurement are subjective and often culturally biased as they are based on self-administered questionnaires or evaluations by other people. Since the VoiceSense analysis is based on prosodic speech patterns, which are common to all humans beyond language and culture, our analysis is objective, language and culture independent. [Scoring all people on one objective scale enables true personalization].
- Fully language independent
- Fully speaker independent (no need for baseline)
- Cloud, local and mobile environments
- Excellent performance (low false alarm, high accuracy)
- Worldwide patented (granted)