Signal processing for multimedia and telecommunications
Focus
Multimedia signal processing, with emphasis on audio and video signals. This includes machine vision; image-based 3D scene modelling; pattern analysis and classification; audio-video analytics; multidimensional signal processing; multimedia forensics; sound analysis, synthesis and processing; computational acoustics; space-time audio processing; spatial audio; acoustic scene analysis and rendering; music information retrieval; and more.
Wireless communication, with focus on cooperative and cognitive wireless systems; multi-carrier/multi-antenna systems (channel estimation, soft-iterative receivers, interference cancellation, channel-aware resource allocation); distributed signal processing for ad-hoc and wireless sensor networks (localization and synchronization); stochastic methods for traffic data aggregation and analysis in intelligent transport systems; home automation; and more.
Wireless communication, with focus on cooperative and cognitive wireless systems; multi-carrier/multi-antenna systems (channel estimation, soft-iterative receivers, interference cancellation, channel-aware resource allocation); distributed signal processing for ad-hoc and wireless sensor networks (localization and synchronization); stochastic methods for traffic data aggregation and analysis in intelligent transport systems; home automation; and more.
Most relevant research achievements
A novel approach to plenacoustic scene analysis and rendering
The ISPG has recently developed an innovative approach to soundfield analysis based on spatially extended arrays of microphones, which jointly employs space-time processing techniques and pattern analysis solutions to analyze and model the plenacoustic function (soundfield along all possible acoustic rays crossing a window of observation). This “soundfield camera” (audio equivalent of the “lightfield camera”) takes “plenacoustic snapshots” of the wavefield, which can be analyzed and processed to extract and reshape the acoustic scene. ISPG also developed plenacoustic projectors, which can be used for rendering a plenacoustic scene.
Interactive soundfield modeling
The ISPG has recently developed an acoustic beam and ray tracer based whose efficiency is capable of supporting applications of interactive soundfield modeling (e.g. virtual walkthrough in architectural acoustics, interactive virtual acoustics in gaming applications, etc.). This method uses an innovative parameterization of acoustic rays to swiftly construct the beam tree (which encodes the whole soundfield in the 3D environment) through a lookup of the visibility data structure. Acoustic rays can also be traced on the fly, through a beam-tree lookup.
No-Reference Pixel Video Quality Monitoring
Video transmitted over an error-prone network can be received with degradations due to packet losses. No-reference quality monitoring algorithms are the most practical way to measure the quality of the received video. Such methods usually rely on the availability of the corrupted bit-stream but this, in some situations, is not possible because the bitstream is encrypted or processed by third-party decoders, and only the decoded pixel values can be used. The major issue in this case is the lack of knowledge on which video regions have been lost. ISPG recently developed an innovative method for MAP estimation of the pattern of lost macro-blocks, which only uses the knowledge of the decoded pixels. This can be used for no-reference video quality monitoring.
Revealing the Traces of JPEG Compression Anti-Forensics
Due to the lossy nature of transform coding, JPEG introduces characteristic traces in the compressed images. A forensic analyst might reveal these traces by analyzing the histogram of DCT coefficients and exploit them to identify tampering/forgery. A knowledgeable adversary, however, could conceal the traces of JPEG compression by adding a dithering noise in the DCT domain. We studied the processing chain in the case of JPEG compression anti-forensics. Taking the perspective of the forensic analyst, we showed how it is possible to counter this type of anti-forensic action by revealing the traces of JPEG compression, regardless of the quantization matrix that is being used. The method has been extended to the case of multiple JPEG compressions and video compression.
Bayesian inference for Intelligent Transportation Systems (ITS)
Tracking of moving vehicles, persons and assets is a fundamental requirement for smart mobility. That of the intelligent transportation system is a new paradigm that relies on sensing and positioning technologies to estimate and control the traffic flows. Achievement of this research is the development of new processing methodologies for the accurate tracking of moving targets and for the reconstruction of the overall traffic field from the sensed data with fixed/mobile devices (e.g., in-car navigators and location-enabled mobile phones). Remarkable is the precise tracking in harsh environments with dense multipath and non-line-of-sight conditions when using short-range wireless networks.
Wireless Cloud Network in industrial monitoring and control
Wireless networking is causing a disruptive revolution in the field of industrial quality and process control as flexible cable-replacing technology. Over the last years we gain experience in the areas of oil/gas exploration, petroleum refineries, railway control, and energy distribution in addressing issues such as reliability and delay control of wireless systems for industrial applications. Wireless Cloud Network (WCN) is the paradigm where messages are flooded from sensors/actuators to a host station (and vice versa) through a self-contained dense network characterized by massively air-interacting “cloud” nodes.
Heterogeneous Networks and small/femto-cells for LTE systems
Cellular traffic is expected to grow three orders of magnitude in the next decade, doubling every year. This growth will only be possible by making LTE base stations pervade the territory and inter-operate with other protocols. This is the essence of the HetNet paradigm. This viral deployment of small cells in next few years requires large-capacity backhauling systems such as cables (optical fiber and copper). We patented and validated a novel network architecture to relay wireless signals over cable (WoC), that integrates into the HetNet the existing last-mile copper to handle multiple small cells as a distributed antennas system. Key-benefit of WoC is flexibility to be integrated in any of the network architectures for large-scale antennas systems with a remarkable (x100) reduction of network-induced CO2 emissions (so called Green-networks).
The ISPG has recently developed an innovative approach to soundfield analysis based on spatially extended arrays of microphones, which jointly employs space-time processing techniques and pattern analysis solutions to analyze and model the plenacoustic function (soundfield along all possible acoustic rays crossing a window of observation). This “soundfield camera” (audio equivalent of the “lightfield camera”) takes “plenacoustic snapshots” of the wavefield, which can be analyzed and processed to extract and reshape the acoustic scene. ISPG also developed plenacoustic projectors, which can be used for rendering a plenacoustic scene.
Interactive soundfield modeling
The ISPG has recently developed an acoustic beam and ray tracer based whose efficiency is capable of supporting applications of interactive soundfield modeling (e.g. virtual walkthrough in architectural acoustics, interactive virtual acoustics in gaming applications, etc.). This method uses an innovative parameterization of acoustic rays to swiftly construct the beam tree (which encodes the whole soundfield in the 3D environment) through a lookup of the visibility data structure. Acoustic rays can also be traced on the fly, through a beam-tree lookup.
No-Reference Pixel Video Quality Monitoring
Video transmitted over an error-prone network can be received with degradations due to packet losses. No-reference quality monitoring algorithms are the most practical way to measure the quality of the received video. Such methods usually rely on the availability of the corrupted bit-stream but this, in some situations, is not possible because the bitstream is encrypted or processed by third-party decoders, and only the decoded pixel values can be used. The major issue in this case is the lack of knowledge on which video regions have been lost. ISPG recently developed an innovative method for MAP estimation of the pattern of lost macro-blocks, which only uses the knowledge of the decoded pixels. This can be used for no-reference video quality monitoring.
Revealing the Traces of JPEG Compression Anti-Forensics
Due to the lossy nature of transform coding, JPEG introduces characteristic traces in the compressed images. A forensic analyst might reveal these traces by analyzing the histogram of DCT coefficients and exploit them to identify tampering/forgery. A knowledgeable adversary, however, could conceal the traces of JPEG compression by adding a dithering noise in the DCT domain. We studied the processing chain in the case of JPEG compression anti-forensics. Taking the perspective of the forensic analyst, we showed how it is possible to counter this type of anti-forensic action by revealing the traces of JPEG compression, regardless of the quantization matrix that is being used. The method has been extended to the case of multiple JPEG compressions and video compression.
Bayesian inference for Intelligent Transportation Systems (ITS)
Tracking of moving vehicles, persons and assets is a fundamental requirement for smart mobility. That of the intelligent transportation system is a new paradigm that relies on sensing and positioning technologies to estimate and control the traffic flows. Achievement of this research is the development of new processing methodologies for the accurate tracking of moving targets and for the reconstruction of the overall traffic field from the sensed data with fixed/mobile devices (e.g., in-car navigators and location-enabled mobile phones). Remarkable is the precise tracking in harsh environments with dense multipath and non-line-of-sight conditions when using short-range wireless networks.
Wireless Cloud Network in industrial monitoring and control
Wireless networking is causing a disruptive revolution in the field of industrial quality and process control as flexible cable-replacing technology. Over the last years we gain experience in the areas of oil/gas exploration, petroleum refineries, railway control, and energy distribution in addressing issues such as reliability and delay control of wireless systems for industrial applications. Wireless Cloud Network (WCN) is the paradigm where messages are flooded from sensors/actuators to a host station (and vice versa) through a self-contained dense network characterized by massively air-interacting “cloud” nodes.
Heterogeneous Networks and small/femto-cells for LTE systems
Cellular traffic is expected to grow three orders of magnitude in the next decade, doubling every year. This growth will only be possible by making LTE base stations pervade the territory and inter-operate with other protocols. This is the essence of the HetNet paradigm. This viral deployment of small cells in next few years requires large-capacity backhauling systems such as cables (optical fiber and copper). We patented and validated a novel network architecture to relay wireless signals over cable (WoC), that integrates into the HetNet the existing last-mile copper to handle multiple small cells as a distributed antennas system. Key-benefit of WoC is flexibility to be integrated in any of the network architectures for large-scale antennas systems with a remarkable (x100) reduction of network-induced CO2 emissions (so called Green-networks).
