Naoko Tosa, an artist of ATR Media Integration and Communications Research Laboratories in Kyoto, Japan, created a computer-aided image of a baby. The baby was named the "neuro-baby" because it took advantage of a neural network with the help of the Institute of Industrial Sciences, Tokyo University. The "neuro-baby" consists of only a head and a robotic arm and can make various facial expressions. What is interesting about the "neuro-baby" is that it can sense what emotions people are feeling according to their voice intonation and strength of a handgrasp.

The baby has been trained to learn the relationships between the characteristics of human voices and the emotions represented by the voices using the neural network. The voice characteristics are measured as the maximum amplitude and the number of zero intersections which are observed in their wave pattern. The emotions are classified as joy, anger, sorrow, and cheerfulness. To the "neuro-baby," soft and high voices are usually recognized as cheerful while loud and low voices are most likely thought to express anger.

The Handshaking Device (HSD) gives the "neuro-baby" another clue to guess the emotional state of the person who communicates with him. According to the combination of the strength of the grip and the intonation of the person's voice, the "neuro-baby" can tell what emotion you are feeling.

WHat issues in cognitive science does the neuro-baby raise? Is this the "brain in a vat"? To what extent can emotional knowledge be computed? Is emotional knowledge determinstically stated or implemented probabilistically?