Voice is a rich and complex channel that can contain several indicators about the speaker beyond spoken words such as their biological sex, age, and even identity.
In order to personalize each experience with our bots, we have designed an algorithm to detect biological sex and age range. Our bots are able to recognize these characteristics in less than two seconds and modulate the conversational path. These characteristics can also be regarded as the first level of fraud detection.
The performance of these algorithms is measured by their classification precision and reaches 98% accuracy for biological sex recognition and 90% for age detection.