With the backdrop laid, Shimada discussed problems they encountered calibrating the handwriting and voice recognition software. Firstly, Nintendo realized that existing voice profiles are strongly skewed towards adults, and while the game was Nintendo's attempt at appealing to an older audience, Nintendo wanted younger players to enjoy the game as well. So Shimada's team had to develop a new profile for children, which required a variety of sample recordings: 130 words read by 20 children … the sons and daughters of NCL employees, of course. They were able to record in Nintendo's professional sound studio for a representative quiet environment, but had to resort to the street outside of NCL for a noisier environment. (Dragging the recording equipment over a long distance would have been very difficult.)

Voice recognition posed two other major challenges. Firstly, determining the end of voice input was a trade-off between accuracy and speed. Increasing the period of silence needed to signal the end of voice input ensures that words or syllables that should be interpreted together are not analyzed separately. However, doing so ruins the pacing by increasing lag, which is far from ideal for a game that encourages speed.

Secondly, when performing focus testing the team discovered that older players had trouble annunciating certain syllables, causing significant trouble for the software (e.g. ki was being detected as chi). To get around the problem, the team included a "sub-dictionary" of alternative pronunciations for a word that would be considered equally correct. Since the game often only expects one among a handful of words to be spoken, they also implemented a ranking system—the voice sample is compared against all words, and if one of the two closest matches is the correct answer, the answer is considered correct.

For handwriting recognition, Shimada knew they would need as much test data as possible and that they would need to use the DS touch screen for sample input. They quickly created software that would record and save user input to flash memory on the DS card for uploading to their database, from which they could perform batch analysis on the sample set to measure the accuracy of the recognition code and settings.

Of course, since the player is being timed in Brain Age, and rushed players tend to write sloppily, the handwriting recognition had to be very robust. It also needed to be very quick for what players were calculating and writing numbers, so they made the input cut-off time the moment the stylus is released from the touch screen.

Brain Age was successfully released on time in Japan and was widely popular. However, with only voice and handwriting recognition tools polished for Japanese, Shimada's job was far from done....