Jeremy Reed

On the importance of modeling temporal information in music tag annotation

Music is an art form in which sounds are organized in time; however, current approaches for determining similarity and classification largely ignore temporal information. This paper presents an approach to automatic tagging which incorporates temporal aspects of music directly into the statistical models, unlike the typical bag-of-frames paradigm in traditional music information retrieval techniques. Vector quantization on song segments leads to a vocabulary of acoustic segment models. An unsupervised, iterative process that cycles between Viterbi decoding and Baum-Welch estimation builds transcripts of this vocabulary. Latent semantic analysis converts the song transcriptions into a vector for subsequent classification using a support vector machine for each tag. Experimental results demonstrate that the proposed approach performs better in 15 of the 18 tags. Further analysis demonstrates an ability to capture local timbral characteristics as well as sequential arrangements of acoustic segment models.

Gaussian multiple access channels for radar and communications spectrum sharing

This paper develops the fundamental performance bounds for a radar operating in the presence of communication interference. To do so, we recast radar as a communication problem with the radar waveform acting as a convolutional encoder, which allows for the description of the achievable information-theoretic limits of the radar problem. While information theory has been applied to communications and radar, there has been little work on applying information theory to systems that explicitly combine sensing and communications. The new viewpoint presented in this paper, which does not require new mathematics, allows one to recast the problem of a radar operating in the presence of communication interference as a Gaussian multiple access channel problem. The derived bounds describe the fundamental limits on the amount of information the radar can obtain about the scattering environment in the presence of communications interference.

Performance bounds for an OFDM-based joint radar and communications system

We investigate a joint radar and communications performance bound for a single pair of bistatic transmit and receive antennas that uses orthogonal frequency-division multiplexing (OFDM) transmission. The OFDM transmission signal is designed to simultaneously meet radar target detection and communications signal recovery objectives using training and information symbols on different subcarriers. Given a frequency-selective Rayleigh-distributed fading channel, equally powered and equally spaced training symbols across the occupied bandwidth minimize the mean-square error of the channel estimate, which is tied to a lower bound on ergodic channel capacity. We perform a hypothesis test on the resulting channel estimate to detect and locate targets. We examine the performance region of capacity versus probability of detection in an example scenario in which a single stationary target is present.

Performance of a joint radar-communication system in doubly-selective channels

When radar and communication systems are co-located and operating simultaneously in the same frequency band, interference can be addressed by designing the systems to share resources. This paper develops a framework for cooperative operation of bistatic radar and wireless communication systems. We adopt a low-complexity linear minimum mean square error (LMMSE) optimal pilot symbol aided modulation (PSAM) scheme, and show the achievability region of the joint radar-communication system designed for simultaneous operation in a wireless channel that is both frequency and time-selective.

Preference Music Ratings Prediction Using Tokenization and Minimum Classification Error Training

In order to address two main limitations of current content-based music recommendation approaches, an ordinal regression algorithm for music recommendation that incorporates dynamic information is presented. Instead of assuming that local spectral features within a song are identically and independently distributed examples of an underlying probability density, music is characterized by a vocabulary of acoustic segment models (ASMs), which are found with an unsupervised process. Further, instead of classifying music based on subjective classes, such as genre, or trying to find a universal notion of similarity, songs are classified based on personal preference ratings. The ordinal regression approach to perform the ratings prediction is based on the discriminative-training algorithm known as minimum classification error (MCE) training. Experimental results indicate that improved temporal modeling leads to superior performance over standard spectral-based music representations. Further, the MCE-based preference ratings algorithm is shown to be superior over two other systems. Analysis demonstrates that the superior performance is due to MCE being a non-conservative algorithm that demonstrates immunity to outliers.

Performance of a Linear-Detector Joint Radar-Communication System in Doubly Selective Channels

When radar and communication systems are colocated and operating simultaneously in the same frequency band, interference can be addressed by designing the systems to share resources. This paper develops a framework for cooperative operation of single-input single-output bistatic radar and wireless communication systems. We adopt a low-complexity linear minimum mean square error (LMMSE) optimal pilot symbol aided modulation scheme, and show the achievability region of the joint radar-communication system designed for simultaneous operation in a wireless channel that is both frequency and time-selective. Since the LMMSE estimate requires accurate knowledge of channel statistics, we give a lower bound on performance using least-squares estimation that operates purely on the received signal. Finally, to better match power levels between systems, we compare the performance using optimal training sequences with a suboptimal scheme using Barker sequences.