The challenge of inverse-E: the RASTA-PLP method

Syllabic (∼2–5 Hz) and fluctuation (∼1–10 Hz) ranges in speech and auditory processing

“…Hermansky and co-workers first employed bandpass filtering in the fluctuation range for improving ASR (Hermansky et al, 1991, Hermansky and Morgan, 1994). Relative immunity to steady background noise was achieved in their ‘RASTA’ system by bandpass filtering the log-envelopes from ~0.26-12.8 Hz: “The key idea here is to suppress constant factors in each spectral component of the short-term auditory-like spectrum…” (Hermansky et al 1991).…”

“…Relative immunity to steady background noise was achieved in their ‘RASTA’ system by bandpass filtering the log-envelopes from ~0.26-12.8 Hz: “The key idea here is to suppress constant factors in each spectral component of the short-term auditory-like spectrum…” (Hermansky et al 1991). With Arai and colleagues (Arai et al, 1996, 1999), this idea was extended to the bandpass filtering of cepstral coefficients (a common representation for ASR) and human perceptual experiments.…”

Robust Speech Recognition System Using Conventional and Hybrid Features of MFCC, LPCC, PLP, RASTA-PLP and Hidden Markov Model Classifier in Noisy Conditions

“…A special band-pass filter was added to each frequency sub-band in traditional PLP algorithm in order to smooth out short-term noise variations and to remove any constant offset in the speech channel. Figure 7 shows the most processes involved in RASTA-PLP which include calculating the critical-band power spectrum as in PLP, transforming spectral amplitude through a compressing static nonlinear transformation, filtering the time trajectory of each transformed spectral component by the band pass filter using Equation (4), transforming the filtered speech via expanding static nonlinear transformations, simulating the power law of hearing, and finally computing an all-pole model of the spectrum, as in the PLP [8].…”

“…In order to find the state sequence that is most likely to have produced an observation sequence, Viterbi algorithm was used to find the optimal scoring path of state sequence as shown in Figure 11. The maximum probability of state sequences was defined in Equation (7), and the optimal scoring path of state sequence selected was calculated using Equation (8).…”

Realce E Reconhecimento De Voz Contínua Em Ambientes Adversos

“…Para lidar com isso, no primeiro caso, foi proposto em [132] uma técnica conhecida como RASTA, que executa uma filtragem passa-banda para eliminar as distorções devido à resposta em frequência do canal, suavizando os atributos da voz ao longo do eixo temporal, removendo as frequências menores que 1Hz e atenuando as frequências maiores que 16Hz, já que segundo [133] a maior parte da informação linguística encontra-se compreendida nas frequências de modulação na faixa de 1 a 16 Hz, suavizando assim a correlação entre as componentes de características por meio de uma filtragem das sequências temporais dos parâmetros espectrais. Essa filtragem pode ser realizada em cada componente no domínio log espectral [134] ou no domínio cepstral [135]. Eliminando essas variações das componentes espectrais em comparação com a faixa típica de variação do sinal de voz, melhora-se significativamente a precisão dos sistemas RAV em condições adversas, já que se conseguem suavizar as trajetórias resultantes dos cepstrum em comparação com as originais.…”