Independent Component Analysis.

This dissertation describes a number of algorithms developed to increase the robustness of automatic speech recognition systems with respect to changes in the environment. These algorithms attempt to improve the recognition accuracy of speech recognition systems when they are trained and tested in different acoustical environments, and when a desk-top microphone (rather than a close-talking microphone) is used for speech input. Without such processing, mismatches between training and testing conditions produce an unacceptable degradation in recognition accuracy.
Two kinds of environmental variability are introduced by the use of desk-top microphones and different training and testing conditions: additive noise and spectral tilt introduced by linear filtering. An important attribute of the novel compensation algorithms described in this thesis is that they provide joint rather than independent compensation for these two types of degradation.
Acoustical compensation is applied in our algorithms as an additive correction in the cepstral domain. This allows a higher degree of integration within SPHINX, the Carnegie Mellon speech recognition system, that uses the cepstrum as its feature vector. Therefore, these algorithms can be implemented very efficiently. Processing in many of these algorithms is based on instantaneous signal-to-noise ratio (SNR), as the appropriate compensation represents a form of noise suppression at low SNRs and spectral equalization at high SNRs.
The compensation vectors for additive noise and spectral transformations are estimated by minimizing the differences between speech feature vectors obtained from a "standard" training corpus of speech and feature vectors that represent the current acoustical environment. In our work this is accomplished by minimizing the distortion of vector-quantized cepstra that are produced by the feature extraction module in SPHINX.
In this dissertation we describe several algorithms including the SNR-Dependent Cepstral Normalization, (SDCN) and the Codeword-Dependent Cepstral Normalization (CDCN). With CDCN, the accuracy of SPHINX when trained on speech recorded with a close-talking microphone and tested on speech recorded with a desk-top microphone is essentially the same obtained when the system is trained and tested on speech from the desk-top microphone.
An algorithm for frequency normalization has also been proposed in which the parameter of the bilinear transformation that is used by the signal-processing stage to produce frequency warping is adjusted for each new speaker and acoustical environment. The optimum value of this parameter is again chosen to minimize the vector-quantization distortion between the standard environment and the current one. In preliminary studies, use of this frequency normalization produced a moderate additional decrease in the observed error rate.

/pdf/acoustical-and-environmental-robustness-in-automatic-speech-3yf73aaxj5.pdf

Acoustical and environmental robustness in automatic speech recognition

Speech enhancement and separation are core problems in audio signal processing, with commercial applications in devices as diverse as mobile phones, conference call systems, hands-free systems, or hearing aids. In addition, they are crucial preprocessing steps for noise-robust automatic speech and speaker recognition. Many devices now have two to eight microphones. The enhancement and separation capabilities offered by these multichannel interfaces are usually greater than those of single-channel interfaces. Research in speech enhancement and separation has followed two convergent paths, starting with microphone array processing and blind source separation, respectively. These communities are now strongly interrelated and routinely borrow ideas from each other. Yet, a comprehensive overview of the common foundations and the differences between these approaches is lacking at present. In this paper, we propose to fill this gap by analyzing a large number of established and recent techniques according to four transverse axes: 1 the acoustic impulse response model, 2 the spatial filter design criterion, 3 the parameter estimation algorithm, and 4 optional postfiltering. We conclude this overview paper by providing a list of software and data resources and by discussing perspectives and future trends in the field.

/pdf/a-consolidated-perspective-on-multimicrophone-speech-qnbaqxmtvq.pdf

A Consolidated Perspective on Multimicrophone Speech Enhancement and Source Separation

When performing speaker diarization on recordings from meetings, multiple microphones of different qualities are usually available and distributed around the meeting room. Although several approaches have been proposed in recent years to take advantage of multiple microphones, they are either too computationally expensive and not easily scalable or they cannot outperform the simpler case of using the best single microphone. In this paper, the use of classic acoustic beamforming techniques is proposed together with several novel algorithms to create a complete frontend for speaker diarization in the meeting room domain. New techniques we are presenting include blind reference-channel selection, two-step time delay of arrival (TDOA) Viterbi postprocessing, and a dynamic output signal weighting algorithm, together with using such TDOA values in the diarization to complement the acoustic information. Tests on speaker diarization show a 25% relative improvement on the test set compared to using a single most centrally located microphone. Additional experimental results show improvements using these techniques in a speech recognition task.

/pdf/acoustic-beamforming-for-speaker-diarization-of-meetings-26ampyixrb.pdf

Acoustic Beamforming for Speaker Diarization of Meetings

This paper proposes a new robust adaptive beamformer applicable to microphone arrays. The proposed beamformer is a generalized sidelobe canceller (GSC) with a new adaptive blocking matrix using coefficient-constrained adaptive filters (CCAFs) and a multiple-input canceller with norm-constrained adaptive filters (NCAFs). The CCAFs minimize leakage of the target-signal into the interference path of the GSC. Each coefficient of the CCAFs is constrained to avoid mistracking. The input signal to all the CCAFs is the output of a fixed beamformer. In the multiple-input canceller, the NCAFs prevent undesirable target-signal cancellation when the target-signal minimization at the blocking matrix is incomplete. The proposed beamformer is shown to be robust to target-direction errors as large as 200 with almost no degradation in interference-reduction performance, and it can be implemented with several microphones. The maximum allowable target-direction error can be specified by the user. Simulated anechoic experiments demonstrate that the proposed beamformer cancels interference by over 30 dB. Simulation with real acoustic data captured in a room with 0.3-s reverberation time shows that the noise is suppressed by 19 dB. In subjective evaluation, the proposed beamformer obtains 3.8 on a five-point mean opinion score scale, which is 1.0 point higher than the conventional robust beamformer.

A robust adaptive beamformer for microphone arrays with a blocking matrix using constrained adaptive filters

The quality of sound pickup in large rooms—such as auditoria, conference rooms, or classrooms—is impaired by reverberation and interfering noise sources. These degradations can be minimized by a transducer system that discriminates against sound arrivals from all directions except for that of the desired source. A two‐dimensional array of microphones can be electronically beam steered to accomplish this directivity. This report gives the theory, design, and implementation of a microprocessor system for automatically steering a two‐dimensional microphone array. The signal‐seeking transducer system is implemented as a dual‐beam, “track‐while‐scan” array. It utilizes signal properties to distinguish between desired speech sources and interfering noise. The complete automatic system has been tested in anechoic and medium‐sized auditorium environments, and its performance is discussed.

Computer-steered microphone arrays for sound transduction in large rooms

A signal processing arrangement is connected to a microphone array to form at least one directable beam sound receiver. The directable beam sound receivers are adapted to receive sounds from predetermined locations in a prescribed environment such as auditorium. Signals representative of prescribed sound features received from the plurality of predetermined locations are generated and one or more of the locations is selected responsive to the sound feature signals. A plurality of directable beam sound receivers may be used to concurrently analyze sound features from the predetermined locations. Alternatively, one directable beam sound receiver may be used to scan the predetermined locations so that the sound feature signals therefrom are compared to sound features from a currently selected location.

Sound location arrangement

A method and apparatus for compressing (in 104) and expanding discrete signals (from 201) is disclosed which comprises two compression methods which interact synergistically. Illustratively, a minimum-redundancy Huffman code, which partially represents a compressed signal, is used in conjunction with a uniquely decodable code which comprises two components. The first component represents its own length, the length of the second component and partially represents the compressed signal and the second component partially represents the compressed signal (from 223).

Method and apparatus for two-component signal compression

A five-channel perceptual audio compression system which encodes five matrixed channels switches among a fourteen encoding modes each utilizing a respective different set of matrixed channels. Six of the modes are for encoding the three front channels and eight of the modes are for encoding the two back channels. A number of the modes have as at least one of their respective matrixed channels a matrixed channel which is a function of a) one of the input channels or a sum or difference of two of them, and b) a predicted value of a).

Multi-channel perceptual audio compression system with encoding mode switching among matrixed channels

A five-channel perceptual audio compression system which encodes five matrixed channels switches among a fourteen encoding modes each utilizing a respective different set of matrixed channels. Six of the modes are for encoding the three front channels and eight of the modes are for encoding the two back channels. The rate of perceptual encoding the matrixed channels is controlled by adjusting individual noise thresholds as a function of a global masking threshold.

James David Johnston

Papers

Computer-steered microphone arrays for sound transduction in large rooms

Sound location arrangement

Method and apparatus for two-component signal compression

Multi-channel perceptual audio compression system with encoding mode switching among matrixed channels

Global masking thresholding for use in perceptual coding