Search

International audience; Music source separation is the task of decomposing music into its constitutive components, e.g., yielding separated stems for the vocals, bass, and drums. Such a separation has many applications ranging from rearranging/repurposing the stems (remixing, repanning, upmixing) to full extraction (karaoke, sample creation, audio restoration). Music separation has a long history of scientific activity as it is known to be a very challenging problem. In recent years, deep learning-based systems-for the first time-yielded high-quality separations that also lead to increased commercial interest. However, until now, no open-source implementation that achieves state-of-the-art results is available. Open-Unmix closes this gap by providing a reference implementation based on deep neural networks. It serves two main purposes. Firstly, to accelerate academic research as Open-Unmix provides implementations for the most popular deep learning frameworks, giving researchers a flexible way to reproduce results. Secondly, we provide a pre-trained model for end users and even artists to try and use source separation. Furthermore, we designed Open-Unmix to be one core component in an open ecosystem on music separation, where we already provide open datasets, software utilities, and open evaluation to foster reproducible research as the basis of future development.

International audience; The Multiple-Input Multiple-Output (MIMO) technique is well known in the field of wireless communications, and has recently been proposed to increase the capacity of in-home Power Line Communication (PLC) networks. MIMO PLC employs the Protective Earth in addition to the classical Line and Neutral wires, to form a multi-sensor transmission channel. By considering the additional reception of Common Mode signals, up to 2 simultaneous transmit ports and 4 receive ports can be considered. In order to develop new signal processing strategies optimally exploiting the MIMO propagation characteristics, and to evaluate their performance, accurate modeling of the transmission channel is necessary. This paper presents a comprehensive statistical model of the in-home MIMO PLC channel based on an extensive measurement data base collected in 6 European countries. The study first analyses the main MIMO PLC channel parameters, including path loss parameters and wideband parameters such as the delay spread, coherence bandwidth and small-scale statistics. A focus is then made on the MIMO correlation matrix, and its modelling using empirical statistical distributions. A fully parameterized channel model is then developed, with a particular emphasis on both the frequency fading structure of the transfer function and the correlation matrix between sub-channels. As a result, the proposed model generates random MIMO PLC channel realizations statistically representative of the experimental observations.

International audience; Despite being a well-established ingredient to many wireless systems, multiple input multiple output (MIMO) signal processing has only recently been considered for broadband power line communications (PLC). Adapting multiple-antenna transmission and reception techniques to a wired medium such as the electrical grid requires solving a number of issues, both regarding the physics of electromagnetic transmission and the optimization of the signal processing strategies. In the last few years, significant steps were made to demonstrate the benefits of MIMO PLC and to develop the necessary hardware. As a result, MIMO PLC has been adopted in several broadband PLC specifications, precisely as part of ITU-T G.hn in Recommendation G.9963, and as part of the industry specification HomePlug AV2, which is backward compatible to IEEE 1901. This article reviews important aspects of MIMO PLC, highlighting its similarities and main differences with classical wireless MIMO. It focuses first on the peculiarities of the electrical grid, with a survey of PLC channel and noise characterization in a MIMO context. It further estimates MIMO PLC channel capacity adhering to the electromagnetic compatibility regulations currently in force. Besides, MIMO signal processing techniques most suited to PLC environments are discussed, allowing for throughput predictions. It is found that eigenbeamforming is the best choice for MIMO PLC: the full spatial diversity gain is achieved for highly attenuated channels and maximum multiplexing gain is achieved for channels with low attenuation by utilizing all spatial streams. It is shown that upgrading from a conventional single input single output (SISO) PLC configuration to a 2 by 2 MIMO configuration the throughput can be more than doubled while coverage is increased. The survey concludes with a review of specific MIMO PLC system implementations in the specifications ITU-T G.9963 and HomePlug AV2.

International audience; The idea of using power lines also for communication purposes has already been around at the beginning of the last century. It is now broadly referred to as power line communications (PLCs). The obvious advantage is the widespread availability of electrical infrastructure, so that theoretically, deployment costs are confined to connecting modems to the existing electrical grid. Power line technologies can be grouped into the following: ultra narrowband (UNB) technologies operating at very low data rate in the ULF band (0.3-3 kHz) or in the upper part of the SLF (30-300 Hz) ; narrowband (NB) technologies operating in the VLF, the LF and in parts of the MF bands (3-500 kHz), and capable of data rates ranging between a few kbit/s and around 500 kbit/s ; and broadband (BB) technologies operating in the MF, HF or VHF bands (1.8-250 MHz) and presenting a PHY rate ranging from several Mbit/s to several hundred Mbit/s. This chapter provides an overview of current NB-PLC and BB-PLC systems.

International audience; Multiple input multiple output (MIMO) systems have been used for many years in the field of wireless communications. The huge increase in coverage and capacity offered by MIMO technology are the key benefits of using multiple sensors at the transmitter and receiver . Recently, MIMO technology has been applied in the context of power line communications (PLCs), with the aim of offering higher channel capacity and therefore larger system coverage, by including the use of the protective earth (PE) wire in addition to the line (L) and neutral (N) wires. SNR values observed in typical PLC scenarios can be much higher than in the case of a classical wireless communication. This high SNR condition is beneficial for MIMO transmission, as it ensures a high capacity gain with respect to single input single output (SISO) transmission, even if the channel presents a high degree of spatial correlation. This chapter is divided into two main sections. First, the PLC channel capacity offered by MIMO technology is analysed in details. The channel capacity provides an upper limit for achievable throughput and does not take system implementation or a particular MIMO scheme into consideration. The achievable throughput for different MIMO schemes in an OFDM system with adaptive modulation is then investigated.

International audience; In January 2012, the HomePlug Alliance published the HomePlug AV2 specification. HomePlug AV2 enables gigabit-class connection speeds by leveraging the existing power line wires while remaining fully interoperable with other technologies for in-home connectivity, such as HomePlug AV, HomePlug Green PHY and IEEE 1901. In this chapter, we highlight the key differentiating HomePlug AV2 features as compared with the HomePlug AV technology. At the PHY layer, HomePlug AV2 includes the following: multiple-input-multiple-output (MIMO) signalling with beamforming ; extended frequency band up to 86 MHz ; efficient notching and power back-off to meet EMC regulations ; and EMC-friendly power boost to optimize the transmit power. At the MAC layer, HomePlug AV2 includes the following: power save mode to improve energy efficiency ; short delimiter and delayed acknowledgements to increase the overall transmission efficiency ; and immediate repeating to expand the coverage. All these features improve the quality of service of HomePlug AV2 power line modems by improving coverage and robustness of communication links. The chapter sections provide technical details for all the previously mentioned features. The chapter begins with a brief overview of HomePlug AV technology and a presentation of the overall HomePlug AV2 system architecture. Next sections detail the new PHY and MAC layers. Finally, the performance gain of HomePlug AV2 over HomePlug AV is summarized.

International audience; This chapter introduces channel and noise characterisation for Power Line Communications. Section 1.2 looks at PLC frequency bands and common topologies as these are possibly the most profound stage setters when characterising PLC channel and noise scenarios. Couplers may be of either inductive or capacitive nature with detailed couplingschemes introduced in Section 1.3. In the sequel, the aim of Section 1.4 is to provide information on measurement equipment and procedures that have been used to generate a plurality of results for various chapters throughout the book. Further, Sections 1.5 and 1.6 introduce the underlying concepts of PLC channel and noise modelling, respectively, and guide the reader to the more detailed chapters on each topic. This chapter is rounded off by an appendix that explains the basics of dual conductor transmission line theory, considered interesting background reading for those new to PLC signal propagation.