Data Analysis

Extracting and understanding human emotions from complex, high-dimensional and heterogeneous biomedical data remains a key challenge on our journey throughout the much evolving healthcare and wellness domains.
Traditional data mining and statistical learning approaches must perform preliminary steps in order to obtain effective and more robust features from the data, and then build prediction or clustering models on top of it.

There are lots of challenges on both steps in a scenario of complicated data and lacking sufficient domain knowledge.

Kenko Technology

The unsupervised pre-training breakthrough, new methods to prevent overfitting, the use of general-purpose graphic processing units to speed-up computations and the development of high-level modules to easily build neural networks allowed deep models to establish as state-of-the-art solutions for various tasks. However, with that being said, there are still many challenges that exists when facing biometric data.

address various challenges that still remain unsolved when facing physiological data by using deep architectures with new and improved methods and tools that enable deep learning to interface with physiological information. Key issues such as:

Domain complexity

Different from other application domains (e.g. image and speech analysis), the problems in biomedicine and healthcare are more complicated. Emotions are highly heterogeneous and there is still no complete knowledge on their effects on our bodies in different kinds of scenarios. Moreover, the number of patients is usually limited in a practical clinical scenario and we cannot ask for as many patients as we want. This is why we are focusing on as many validations in different occupational scenarios and scientific research as well, so we can tackle as much real world use cases as we can.

Data volume

Deep learning refers to a set of highly intensive computational models, where tons of network parameters need to be estimated properly. The basis to achieve this goal is the availability of huge amount of data. The healthcare domain specifically lack in its available public data. Consequently, from a big data perspective, the amount of medical data that is needed to train an effective and robust emotional deep learning model would be much more comparing with other media. By forming partnerships with academic institutes, research institutes and hospitals all around the world, we are able to gain access to data volumes that have not been seen before.
Learn more about our experiments

Data quality

Unlike other domains where the data are clean and well-structured, physiological  data is highly heterogeneous, ambiguous, noisy and incomplete. Training a good deep learning models with such massive and variegate data sets is challenging and needs to consider several issues, such as data sparsity, redundancy and missing values. As part of our research, we are addressing issues such as motion artifacts, high signal noise ratio, sensor-specific related issues and more.


Although deep learning models have been successful in quite a few application domains, they are often treated as black boxes. While this might not be a problem in other more deterministic domains such as image annotation (because the end user can objectively validate the tags assigned to the images), in healthcare, not only the quantitative algorithmic performance is important, but also the reason why the algorithms work is relevant. Our unique experiment methodologies allows us to validate both our algorithms result and at the same time to understand the how and why of our algorithms.

Model privacy

Privacy is an important concern in scaling up deep learning (e.g. through cloud computing services). Preserving the privacy of deep learning models is even more challenging, as there are more parameters to be safeguarded. We insist on keeping our users data safe, by complying with different data privacy standards and regulations.

The latest advances of deep learning technologies provide new effective paradigms to obtain end-to-end learning models from complex data.
Deep learning approaches are the vehicle for translating big biomedical data into an improved healthy life and better productivity in both our work environment and personal life.
Given its demonstrated performance in different domains and the rapid progresses of methodological improvements, deep learning introduce exciting new opportunities that Kenko capitalizes.

Machine learning is a general-purpose method of artificial intelligence that can learn relationships from data without the need to define them a priori. Conventional techniques are composed of a single, often linear, transformation of the input space and are limited in their ability to process natural data in their raw form.

Deep learning is different from traditional machine learning in how representations are learned from the raw data. In fact, deep learning allows computational models that are composed of multiple processing layers based on neural networks to learn representations of data with multiple levels of abstraction. The major differences between deep learning and traditional artificial neural networks (ANNs) are the number of hidden layers, their connections and the capability to learn meaningful abstractions of the inputs.

The key aspect of deep learning is that the layers of features are not designed by human engineers, but they are learned from data using a general purpose learning procedure, and it is one of the keys for translation physiological data in emotions and mental states (figure 1 illustrates the differences at a high level).

Figure 1. An example for the difference between an ANN model and Deep Learning model

Figure 1. An example for the difference between
an ANN model and Deep Learning model


The way we process and analyze data has evolved significantly over the past few years. Nowadays, the usage of complex deep learning networks is bringing us closer to understanding our emotions and how they affect our body. Still, there are many obstacles to overcome.

The Challenge

Deep Learning as a game changer

Deep Learning vs traditional methods

Data Analysis

Science   >