WP3 : System Integration

WP objectives

The objectives of this WP will be to optimally integrate the robust spoken dialog into the two platforms considered in the project :

• A handheld devices platform referred to as the "mobile platform" meaning that it has to deal with requirements associated to mobile equipment (limited processing, memory, interface, etc).

• A cockpit simulator that is referred as the "fixed platform", less constrained than the mobile type platform and where processing power is bigger and multi microphone sound capture can be envisaged.

The integration phase of the project sets the stage for the evaluation phase that will be conducted at the end of the project. The objectives of the works conducted in this WP are:

• Take into account the requirements of each application in order to define the optimal recognition system able to support live experiments (real time).

• Definition of the speech recognition interfaces (with users and with other subsystems of the platforms)

• Software and hardware integration of the recognition components into the platforms.

WP duration

The Work package will begin at month 12 and end at month 30. Its starting event will be the completion of the baseline experiments (M1 at month 10).

WP milestones

The main milestones of the WPs are:

M3.1, m15 Initial specification of system configurations

M3.2, m24 Functional integration on both platforms completed

M3.3, m27 Final specification of system configurations

WP tasks

The WP is divided in two main tasks :

Task 1 : fixed platform integration

The cockpit simulator has the following characteristics:

• Study cockpit with a representative display system (four screens) and the capability to display on a projection screen the outside landscape.

• Capability to simulate some scenarios representative of future aeronautic scenario.

• No simulation of engine management and piloting task. The realism of the simulation is limited to the display layer and its interface with systems : (radio management, data link communication, navigation display, etc).

The work identified for this task will be :

• First to identify the requested resources for the speech recognition system. The baseline experiments will provide a first assessment of the impact on recognition of various processing methods. This will be used to issue a first specification of the requested resources for an efficient physical implementation of the whole vocal interaction (sensor & processing).

• Second, definition of the users interface. The simulator will comprise a display layout that is the simplified version of a future commercial aircraft. Previous studies have allow the collection of system and operational requirements on the introduction of the speech recognition in cockpits. The HIWIRE project will capitalise on these results for the global vocal interaction prototyping. New reflection will be raised within our project in order to assess the possibility to introduce a tolerant dialog module. From what we know two antagonists view are expressed by pilots. On the one hand, interviews underlined that it is mandatory to reach a high level of performances (above 98% WER), but on the other hand, studies have shown that when using standardised language, for traffic control communications on the radio channel, they often issue "incorrect" or approximate utterances. As the recognition is expected to be very robust for large vocabulary, it may allow us to meet both constraints (performances and flexibility).

• Third, realise the software integration so that the recognition system will be able to communicate with the simulator subsystems. Due to the fact that this simulator is a "light" study simulator we will minimise costs for this task.

Task 2 : Mobile platform integration

As a contrast to the fixed installation being investigated under task 1, this task will focus on a mobile installation that will be limited in terms of both computational resources and sensor inputs and outputs. It is intended that the best and most appropriate techniques emerging from WP's 1 and 2 will be integrated into an ASR engine running on a mobile platform, and the performance tradeoffs investigated under WP4. The main activities under this task will be:

• to define a suitable application for a mobile platform in an aeronautical environment (eg. hands-free eyes-free access to instructional and test material for aircraft system maintenance engineers)

• to determine the requirements of the chosen application type

• to characterise the constraints imposed by the aeronautic environment on

• the mobile device

• to prototype the ASR design on a conventional platform

• to port the application design to the mobile platform