Overall Objectives
Scientific Foundations
Application Domains
New Results
Contracts and Grants with Industry
Other Grants and Activities

Section: Scientific Foundations

Interfacing artificial and natural parts through neuroprosthetic devices

To overcome the limitations of the present FES centralized architecture, a new FES architecture was proposed according to the SENIS (Stimulation Electrique Neurale dIStribuée) concept: the distribution of i) the stimulation unit with its control near its activator, i.e. its associated neural electrode ii) the implanted sensor with its embedded signal processing.

FES will be thus performed by means of distributed small stimulation units which are driven by an external controller in charge of the coordination of stimulation sequences. Each stimulation unit (called DSU, Distributed Stimulation Unit) will be in charge of the execution of the stimulation pattern, applied to the muscle by means of a neural multipolar electrode. A DSU is composed of analogue and digital parts (§ 6.2.1 ).

The SENIS architecture therefore relies on a set of DSU which communicates with an external controller. We therefore studied the communication architecture and defined an adequate protocol, assuming firstly that the communication should be performed on a wireless medium and secondly that this architecture can also contain distributed measurement units (DMU for sensors).


We mainly focus on implanted devices interfaced with neural structures. Both the knowledge about how to accurately activate neural structures (neurophysiology), and technology including both electrode manufacturing and micro electronics will be studied. Complex electrode geometries, complex stimulus waveforms, and the multiplicity of the implantation sites are the subjects we deal with in order to obtain a selective, progressive and flexible activation of neural structures. Our theoretical approaches are based on:


The development of a closed-loop controller implies the use of sensors whose choice and number are highly constrained by practical, psychological and cosmetic considerations: the stimulation system has to be implanted in order to simplify its use by the patient; it is therefore not possible to cover the person with various external apparatuses. An alternative to artificial sensors is the use of natural sensors already present, which are intact and active below the lesion in the spinal cord of the injured patients. DEMAR is then interested in implanted sensors in order to design complete implanted solutions (stimulation and sensing). As regards sensing, two kinds of sensors will be studied:

In both cases, advanced signal processing applied to biosignals is needed to extract relevant pieces of information.

Patient interface

The patient interacts with the system in three ways:

It's not trivial to integrate all these events in the system. This field of research can learn from tele-operation and Human Machine Interfaces research fields. The patient needs also to get pieces of information of the current state of the system. Sensory feedback have to be implemented in the system such as screen, sound, tactile vibrations, electrical stimulation,... Choosing meaningful pieces of information such as heel contact, and the way to encode it, will be addressed.

Supervision & networking

Activating the system through stimulators, sensors, and analysing patient behaviors need multiple devices that communicate and demand energy. Interfacing natural and artificial parts imply to address problems such as networking, data transfer, energy storage and transfer through wireless links. On such a complex system, supervision is necessary to ensure security at the different involved levels. Fault tolerance and reflex behavior of the system will be studied to improve system reliability particularly when the patient uses it at home without any medical person support. The theoretical approach is based on Petri Nets to design and then analyse the behavior of the entire distributed system. More technological aspects related to RF transmission will be studied.


Logo Inria