Milestone MS7

MS7: Second version of Nao prototype with improved algorithms for audition and interaction
WP number: 5 (with input from WPs 1, 2, 3, 4)
Lead beneficiary: ALD
Delivery date: 31.08.2016 (M32)

Description: The modules defining MS6 are integrated into the Nao robot platform. Embedded
computing capabilities of the robot itself will be complemented by computing power of
workstations linked to the robot as far as necessary and possible.

Verification: Enhanced robot audition and extended HRI for Nao can be demonstrated. Criteria for
enhancement include precision of source localization and speed of tracking performance; speech
recognition rate in noisy and reverberant environments, with competing speech and other
acoustic sources; richness of HRI.

Report: EARS_Report_on_MS7_20160831_final_RG_HL

Milestone MS6

MS6: Hard- and software for final prototype
WP number: 2, 3, 4, 5
Lead beneficiary: FAU
Delivery date: 30.06.2016 (M30)

Description: Microphone array hardware (T1.1, T1.2), software modules for acoustic map
(T2.1,T2.3), focusing (T2.2) and spatial filtering (T2.4,T2.6) for one moving target source,
acoustic echo cancellation (T2.5), audio-visual localisation module (T3.2), interaction behaviour
(T4.1,T4.2,T4.3,T4.5) are ready for integration into the final demonstrator for the ‘Welcome
robot’ scenario. Individual algorithmic performance of the modules is documented.

Verification: By delivery of hardware and software including documentation to WP5.2, 5.3.

Report: EARS_Report_on_MS6_20160630_final_HL

The corresponding software for the final prototype is provided by a repository (for EARS members) and the current version (June 30, 2016) can also be download as zip-archive. The corresponding hardware (Benchmark II head) has been presented at the Review Meeting on April 7, 2016 in Berlin.

Milestone MS5

MS5: First version of Nao prototype with new algorithms for audition and interaction
WP number: 5 (with input from WPs 1, 2, 3, 4)
Lead beneficiary: ALD
Delivery date: 23.12.2015

Description: Based on the microphone arrays of MS4, robust versions with acceptable complexity
of the algorithms of WP2, WP3 and WP4 are incorporated into the available hardware platform
of Nao. This includes first versions of the algorithms resulting from T2.1-T2.7, T3.2, T4.2, T4.4.
Verification by demonstration; benchmarking relative to existing Nao robot audition, with
respect to source localization and tracking performance, and speech recognition rate in noisy
and reverberant environments, with competing speech and other acoustic sources.

Verification: First version of Nao prototype can be demonstrated to track a single desired target and
perform speech recognition in the targeted noisy and reverberant ‘welcome robot scenario’.

Report on MS5: EARS_Report_on_Milestone_MS5_20151223_final_HL_RG

Milestone MS4

MS4:  Microphone array designs for first prototype
WP number: 1 and 5
Lead beneficiary: BGU
Delivery date: 30.06.2015

Description: Based on intermediate results of T1.1, T1.2 an anthropomorphic array deduced
from the existing head with minor modifications and a first robomorphic array configuration are
specified; calibrated sensors provide synchronized low-noise sensor signals (>60 dB SNR full-
scale) digitized and available to signal processing algorithms.

Verification: Specifications, i.e., array geometries, sensor characteristics (directivity, noise sensitivity)
are documented and published to the partners.

Report on MS4: EARS_WP1_MS4_Microphone_array_designs_for_first_prototype_20150630_final_VT


Milestone MS3

MS3:  Robot audition algorithms in generic EARS scenario
WP number: 1, 2, and 5
Lead beneficiary: IMPERIAL
Delivery date: 31.12.2014

Description: For the generic EARS usage scenario (described in D5.2), first results will be shown
for the robot audition algorithms developed in WP2. Real-world data collections recorded using
generic microphone arrays, existing robot microphone arrays, and first anthropomorphic and
robomorphic microphone array designs (from T1.1, T1.2) will form the basis for simulating
algorithms for source localization (T2.1), focusing (T2.2.) and spatial filtering (T2.4). Evaluation
will be carried out regarding the localization accuracy as a function of observation time and the
amount of interference and noise suppression. Moreover, speech recognition scores will be used
to assess the performance of the algorithms.

Verification: Presentation at review meeting; intermediate internal report as a contribution to

Report on MS3: EARS_WP2_MS3_Robotic_Audition_Algorithms_in_Generic_EARS_Scenario_20141229_Final_PN

Milestone MS2

MS2: Internal and public web sites and software portal online
WP number: 6 and 7
Lead beneficiary: FAU
Delivery date: 31.06.2014

Description: As described for Task T6.3., the internal website provides read/write access to
repositories for data, code and documents; templates for documents and code are provided;
mechanisms for secured exchange of data and code between partners are implemented. In
accordance with T 6.1, the (improved) public web site allows read/write access to repositories
for data (audio and video data collections), open source code and documents (publications,
report); links to robot community are installed.

Verification: Relevant information can be accessed both for internal and external use.

Milestone MS1

MS1: Software interface for audio signal acquisition
WP number: 5
Lead beneficiary: ALD,
Delivery date: 31.06.2014

Description: As described for Task T5.1, the Naoqi software platform is modified to allow easy
and direct access to a larger number of microphone signals, while at the same time allowing
reproduction of sound signals by the robot.

Verification: Audio signals in single- and multichannel form can be exchanged between the Nao
hardware platform and the subsequent signal processing software. (See also Deliverable D5.1 for further details)