Yasue Mitsukura

Professor | Keio University

Building a Real-time Emotion Detector from Simple EEG Readings Plus Noise Reduction

Dr. Yasue Mitsukura has developed a system that uses simple devices to measure the emotions people feel while doing everyday things, including eating and performing different tasks. The tool could be a boon for companies looking to determine how potential customers feel about their products.


Dr. Mitsukura, a professor at Keio University in Japan, gave a presentation outlining her work at Upgrade 2020,  the NTT Research Summit. She has developed a technique called the KANSEI model, which uses simple EEG devices to measure brainwaves and determine what a subject is feeling, including stress, interest, sleepiness, “like” or pleasure, and concentration.



Her work involves calculating the signal-to-noise ratio from EEG readings and removing the noise to ultimately determine what the subject was feeling at any point in time.


Virtually any sort of EEG device that can produce data will work with her system. “Our skill is signal processing and if we’ve got a lot of data obtained from the device, the same result can be shown no matter what device is used,” she said. Her team examined more than 50 EEG devices and found they all work well with the KANSEI model.


Most EEG devices involve some type of headband, to which a smart phone is attached. The phone’s camera is used to track what the wearer is seeing, and to collect data.  The model then calculates the KANSEI reading in real time, matches it to what the wearer is seeing at any given moment, and determines how the subject is reacting to various stimuli.


Dr. Mitsukura showed a demonstration video with a young woman wearing the system, the productized version of which is called Neurocam. As she moves about, whenever her KANSEI reading is higher than 60%, it triggers the camera to start filming. The video shows her in various ordinary situations – eating a donut, looking at a cute dog, window shopping, meeting a young man – while the system calculates her KANSAI score and determines what she’s feeling at various moments. She clearly liked the donut and the dog, for example, but especially the young man, who scored a 99.


Dr. Mitsukura gave a couple of examples of how the KANSEI Analyzer may be used in practice. One involved a major Japanese restaurant chain that was testing a new menu item. As a subject samples the item, a cheese souffle topped with kiwi and other fruit, the analyzer records her reactions. At first, her “like” level is 60, which shows she likes the taste, then it soars to 80. But when she tries to scoop up the fruit, her stress level hits 90. The issue: it was difficult to scoop the fruit with the small spoon she was given. The solution: serve it with a fork instead.


Another example involved a printing and design firm that designs direct mail and credit card forms. It is testing a new credit card application form and using the analyzer to determine how well it’s designed. As a subject attempts to fill out the form, the analyzer records her stress levels. In one section, it registers quite high, because the line was too dense and difficult to read. So, the company split the line in two.


For the full transcript of Yasue Mitsukura’s presentation, click here.


Watch Yasue Mitsukura’s full presentation below.

Real Time Emotion Detection Using EEG With Real Time Noise Reduction

Yasue Mitsukura head shot

Yasue Mitsukura
Professor | Keio University