FPGA implementation of dual-microphone delay-and-sum beamforming for in-car speech enhancement and recognition

In an automotive environment, the performance of a speech recognition system is affected by environmental noise if the speech signal is acquired directly from a microphone. Speech enhancement techniques are therefore necessary to improve the speech recognition performance. In this paper, a field-programmable gate array (FPGA) implementation of dual-microphone delay-and-sum beamforming (DASB) for speech enhancement is presented. As the first step towards a cost-effective solution, the implementation described in this paper uses a relatively high-end FPGA device to facilitate the verification of various design strategies and parameters. Experimental results show that the proposed design can produce output waveforms close to those generated by a theoretical (floating-point) model with modest usage of FPGA resources. Speech recognition experiments are also conducted on enhanced in-car speech waveforms produced by the FPGA in order to compare recognition performance with the floating-point representation running on a PC.

Clustering of ad-hoc microphone arrays for robust blind beamforming

This paper proposes a clustered approach for blind beamfoming from ad-hoc microphone arrays. In such arrangements, microphone placement is arbitrary and the speaker may be close to one, all or a subset of microphones at a given time. Practical issues with such a configuration mean that some microphones might be better discarded due to poor input signal to noise ratio (SNR) or undesirable spatial aliasing effects from large inter-element spacings when beamforming. Large inter-microphone spacings may also lead to inaccuracies in delay estimation during blind beamforming. In such situations, using a cluster of microphones (ie, a sub-array), closely located both to each other and to the desired speech source, may provide more robust enhancement than the full array. This paper proposes a method for blind clustering of microphones based on the magnitude square coherence function, and evaluates the method on a database recorded using various ad-hoc microphone arrangements.

Clustered blind beamforming from ad-hoc microphone arrays

Microphone arrays have been used in various applications to capture conversations, such as in meetings and teleconferences. In many cases, the microphone and likely source locations are known \emph{a priori}, and calculating beamforming filters is therefore straightforward. In ad-hoc situations, however, when the microphones have not been systematically positioned, this information is not available and beamforming must be achieved blindly. In achieving this, a commonly neglected issue is whether it is optimal to use all of the available microphones, or only an advantageous subset of these. This paper commences by reviewing different approaches to blind beamforming, characterising them by the way they estimate the signal propagation vector and the spatial coherence of noise in the absence of prior knowledge of microphone and speaker locations. Following this, a novel clustered approach to blind beamforming is motivated and developed. Without using any prior geometrical information, microphones are first grouped into localised clusters, which are then ranked according to their relative distance from a speaker. Beamforming is then performed using either the closest microphone cluster, or a weighted combination of clusters. The clustered algorithms are compared to the full set of microphones in experiments on a database recorded on different ad-hoc array geometries. These experiments evaluate the methods in terms of signal enhancement as well as performance on a large vocabulary speech recognition task.

Microphone array beamforming approach to blind speech separation

In this paper, we present a microphone array beamforming approach to blind speech separation. Unlike previous beamforming approaches, our system does not require a-priori knowledge of the microphone placement and speaker location, making the system directly comparable other blind source separation methods which require no prior knowledge of recording conditions. Microphone location is automatically estimated using an assumed noise field model, and speaker locations are estimated using cross correlation based methods. The system is evaluated on the data provided for the PASCAL Speech Separation Challenge 2 (SSC2), achieving a word error rate of 58% on the evaluation set.

A superdirective 3-element array for adaptive beamforming

A small array composed of three monopole elements with very small element spacing on the order of λ/6 to λ/20 is considered for application in adaptive beamforming. The properties of this 3-port array are governed by strong mutual coupling. It is shown that for signal-to-noise maximization, it is not sufficient to adjust the weights to compensate for the effects of mutual coupling. The necessity for a RF-decoupling network (RF-DN) and its simple realization are shown. The array with closely spaced elements together with the RF-DN represents a superdirective antenna with a directivity of more than 10 dBi. It is shown that the required fractional frequency bandwidth and the available unloaded Q of the antenna and RF-DN structure determine the lower limit for the element spacing.

On power minimisation and SNR maximisation of distributed beamforming in cooperative communication systems

A comparison of relay power minimisation subject to received signal-to-noise ratio (SNR) at the receiver and SNR maximisation subject to the total transmitted power of relays for a typical wireless network with distributed beamforming is presented. It is desirable to maximise receiver quality-of-service (QoS) and also to minimise the cost of transmission in terms of power. Hence, these two optimisation problems are very common and have been addressed separately in the literature. It is shown that SNR maximisation subject to power constraint and power minimisation subject to SNR constraint yield the same results for a typical wireless network. It proves that either one of the optimisation approaches is sufficient.

Analysis of distributed beamforming in cooperative communications network with phase shifter based smart antenna nodes

This research study investigates the application of phase shifter-based smart antenna system in distributed beamforming. It examines the way to optimise the transmit power by jointly maximising the directivity of the array antennas and the weight vector for distributed beamforming. This research study concludes that maximising directivity can lead to better transmit power minimisation compared to maximising field intensity. This study also concludes that signal to noise power ratio maximisation subject to a power constraint and power minimisation subject to a signal to noise power ratio constraint yield the same results.

Robust Semi-Blind Estimation for Beamforming Based MIMO Wireless Communication

In this paper, we present robust semi-blind (SB) algorithms for the estimation of beamforming vectors for multiple-input multiple-output wireless communication. The transmitted symbol block is assumed to comprise of a known sequence of training (pilot) symbols followed by information bearing blind (unknown) data symbols. Analytical expressions are derived for the robust SB estimators of the MIMO receive and transmit beamforming vectors. These robust SB estimators employ a preliminary estimate obtained from the pilot symbol sequence and leverage the second-order statistical information from the blind data symbols. We employ the theory of Lagrangian duality to derive the robust estimate of the receive beamforming vector by maximizing an inner product, while constraining the channel estimate to lie in a confidence sphere centered at the initial pilot estimate. Two different schemes are then proposed for computing the robust estimate of the MIMO transmit beamforming vector. Simulation results presented in the end illustrate the superior performance of the robust SB estimators.

Design and analysis of integrated optic waveguide delay line for beamforming applications

This paper presents a new approach for Optical Beam steering using 1-D linear arrays of curved wave guides as delay line. The basic structure for generating delay is the curved/bent waveguide and hence its Analytical modelling involves evaluation of mode profiles, propagation constants and losses become important. This was done by solving the dispersion equation of a bent waveguide with specific refractive index profiles. The phase shifts due to S-bends are obtained and results are compared with theoretical values. Simulations in 2-D are done using BPM and Matlab.

Channel training signal design for reciprocal multiple antenna systems with beamforming

Fast and efficient channel estimation is key to achieving high data rate performance in mobile and vehicular communication systems, where the channel is fast time-varying. To this end, this work proposes and optimizes channel-dependent training schemes for reciprocal Multiple-Input Multiple-Output (MIMO) channels with beamforming (BF) at the transmitter and receiver. First, assuming that Channel State Information (CSI) is available at the receiver, a channel-dependent Reverse Channel Training (RCT) signal is proposed that enables efficient estimation of the BF vector at the transmitter with a minimum training duration of only one symbol. In contrast, conventional orthogonal training requires a minimum training duration equal to the number of receive antennas. A tight approximation to the capacity lower bound on the system is derived, which is used as a performance metric to optimize the parameters of the RCT. Next, assuming that CSI is available at the transmitter, a channel-dependent forward-link training signal is proposed and its power and duration are optimized with respect to an approximate capacity lower bound. Monte Carlo simulations illustrate the significant performance improvement offered by the proposed channel-dependent training schemes over the existing channel-agnostic orthogonal training schemes.

Decode-and-forward relay beamforming for secrecy with finite-alphabet input

In this letter, we compute the secrecy rate of decode-and-forward (DF) relay beamforming with finite input alphabet of size M. Source and relays operate under a total power constraint. First, we observe that the secrecy rate with finite-alphabet input can go to zero as the total power increases, when we use the source power and the relay weights obtained assuming Gaussian input. This is because the capacity of an eavesdropper can approach the finite-alphabet capacity of 1/2 log(2) M with increasing total power, due to the inability to completely null in the direction of the eavesdropper. We then propose a transmit power control scheme where the optimum source power and relay weights are obtained by carrying out transmit power (source power plus relay power) control on DF with Gaussian input using semi-definite programming, and then obtaining the corresponding source power and relay weights which maximize the secrecy rate for DF with finite-alphabet input. The proposed power control scheme is shown to achieve increasing secrecy rates with increasing total power with a saturation behavior at high total powers.

Decode-and-forward relay beamforming for secrecy with imperfect CSI and multiple eavesdroppers

In this paper, we evaluate secrecy rates in cooperative relay beamforming in the presence of imperfect channel state information (CSI) and multiple eavesdroppers. A source-destination pair aided by.. out of.. relays, 1 <= k <= M, using decode-and-forward relay beamforming is considered. We compute the worst case secrecy rate with imperfect CSI in the presence of multiple eavesdroppers, where the number of eavesdroppers can be more than the number of relays. We solve the optimization problem for all possible relay combinations to find the secrecy rate and optimum source and relay weights subject to a total power constraint. We relax the rank-1 constraint on the complex semi-definite relay weight matrix and use S-procedure to reformulate the optimization problem that can be solved using convex semi-definite programming.