Itakura-Saito Measure Research Articles

Learning HMM State Sequences from Phonemes for Speech Synthesis

This paper presents a technique for learning hidden Markov model (HMM) state sequences from phonemes, that combined with modified discrete cosine transform (MDCT), is useful for speech synthesis. Mel-cepstral spectral parameters, currently adopted in the conventional methods as features for HMM acoustic modeling, do not ensure direct speech waveforms reconstruction. In contrast to these approaches, we use an analysis/synthesis technique based on MDCT that guarantees a perfect reconstruction of the signal frame feature vectors and allows for a 50% overlap between frames without increasing the data rate. Experimental results show that the spectrograms achieved with the suggested technique behave very closely to the original spectrograms, and the quality of synthesized speech is conveniently evaluated using the well known Itakura-Saito measure.

KEMAR 마네킹을 이용한 단이 보청기용 FDSI 빔포밍 알고리즘의 정량적 평가

Abstract: To enhance the speech perception of hearing aid users in noisy environment, most hearing aid devicesadopt various beamforming algorithms such as the first-order differential microphone (DM1) and the two-stage direc-tional microphone (DM2) algorithms that maintain sounds from the direction of the interlocutor and reduce the ambi-ent sounds from the other directions. However, these conventional algorithms represent poor directionality abilityin low frequency area. Therefore, to enhance the speech perception of hearing aid uses in low frequency range, ourgroup had suggested a fractional delay subtraction and integration (FDSI) algorithm and estimated its theoretical per-formance using computer simulation in previous article. In this study, we performed a KEMAR test in non-rever-berant room that compares the performance of DM1, DM2, broadband beamforming (BBF), and proposed FDSIalgorithms using several objective indices such as a signal-to-noise ratio (SNR) improvement, a segmental SNR (seg-SNR) improvement, a perceptual evaluation of speech quality (PESQ), and an Itakura-Saito measure (IS). Experi-mental results showed that the performance of the FDSI algorithm was −3.26-7.16 dB in SNR improvement, −1.94-5.41 dB in segSNR improvement, 1.49-2.79 in PESQ, and 0.79-3.59 in IS, which demonstrated that the FDSI algo-rithm showed the highest improvement of SNR and segSNR, and the lowest IS. We believe that the proposed FDSIalgorithm has a potential as a beamformer for digital hearing aid devices.Key words: hearing aids, directional microphone, beamforming

A Combined Psychoacoustic Approach for Stereo Acoustic Echo Cancellation

Acoustic echo cancelers (AECs) are used in teleconferencing systems to reduce undesired echoes originating from coupling between the loudspeaker and the microphone. Better performance in terms of sound localization can be achieved only by using multichannel AECs. By the use of a two-channel (stereo) system, it is already possible to obtain more realistic performance than the monochannel case because listeners have more spatial information that helps to identify the speaker position. In this paper, a novel approach to stereo AEC is proposed based on a combination of psychoacoustic effects. The principle of the proposed structure is to use an approach based on the “missing-fundamental” phenomenon at low frequencies (below 500 Hz) and a modified phase modulation approach at higher frequencies. Several results are presented in terms of magnitude-squared coherence, Itakura-Saito measure, convergence speed of adaptive filters, interaural cross-correlation, and subjective evaluation, in order to confirm the validity of the proposed approach.

Text-directed speech enhancement employing phone class parsing and feature map constrained vector quantization

There are many situations where non-real-time speech enhancement is required. For such applications, employing any available a priori knowledge can lead to more effective enhancement solutions. In this study, a novel text-directed speech enhancement algorithm is developed for usage in non-real-time applications. In our approach, the text of the intended dialogue is used to partition noisy speech into regions of broad phoneme classifications. Classes considered include stops, fricatives, affricates, nasals, vowels, semivowels, diphthongs and silence. These partitions are then used to direct a new vector quantizer based enhancement scheme in which phone-class directed constraints are applied to improve speech quality. The proposed algorithm is evaluated using both objective as well as subjective quality assessment techniques. It is shown that the text-directed approach improves the quality of the degraded speech over a broad range of noise sources (i.e., flat communications channel noise, aircraft cockpit noise, helicopter fly-by noise, and automobile highway noise) and over a broad range of signal-to-noise ratios (i.e., 10, 5, 0 and −5 dB). In each case, the proposed method is shown consistently to exhibit improved objective quality over linear and generalized spectral subtraction, as well as the Auto-LSP constrained iterative enhancement method using the Itakura-Saito measure and a 100-sentence evaluation speech corpus. Subjective quality assessment was conducted in the form of an A-B comparison test. Results of these evaluations demonstrate that, for wideband noise distortions, the proposed algorithm is preferred over the unprocessed noisy speech more than 2 to 1, while the proposed algorithm is preferred over spectral subtraction by more than 3 to 1.

A frequency weighted Itakura-Saito spectral distance measure

A new spectral distance measure is defined by inserting a multiplicative frequency weighting term into the conventional Itakura-Saito measure. When the weighting function is a simple one pole function, then the minimization of the distance between a signal spectrum and an arbitrary N-pole filter results in a set of linear equations that is symmetric and solvable by Cholesky decomposition. When the weighting function is a multiple pole function, the resulting spectral distance minimization produces a set of nonlinear algebraic equations, but fortunately a simple method exists for obtaining an approximate solution which can be refined using the Newton-Raphson method. Results of some preliminary trials in applying the technique to LPC vocoding are described.