Multimicrophone Speech Dereverberation: Experimental Validation

Modulation Spectral Features for Robust Far-Field Speaker Identification

“…In turn, the ITU-T G.191 tool is used to convolve room impulse responses measured from an office environment with clean speech signals. The measured room impulse responses used in our experiments are described in [20] and were collected with a six-channel microphone array and corresponded to s. Microphones were omnidirectional and spaced 5 cm apart in a linear array. The speaker was placed at a 90 angle with respect to the center of the array at a distance of 94 cm.…”

“…Each speaker reads four short fables and utters 33 short sentences; the latter For the baseline we experiment with three signal level speech enhancement schemes in combination with CMSVN. The selected multichannel dereverberation algorithms are the ones that showed superior performance in the automatic speech recognition test described in [20]. The dereverberation algorithms include a delay-and-sum beamformer (DSB), the multichannel cepstrum based algorithm described in [37], and a frequency domain subspace-based algorithm [38]; more detail regarding the speech enhancement algorithms can be found in [20].…”

“…The selected multichannel dereverberation algorithms are the ones that showed superior performance in the automatic speech recognition test described in [20]. The dereverberation algorithms include a delay-and-sum beamformer (DSB), the multichannel cepstrum based algorithm described in [37], and a frequency domain subspace-based algorithm [38]; more detail regarding the speech enhancement algorithms can be found in [20]. Table III shows performance figures attained with the systems for the multichannel reverberant speech signals.…”

Using neural network front-ends on far field multiple microphones based speech recognition

“…Typically, multiple channel data is enhanced before recognition. The most representative multi-channel signal enhancement methods are beamforming techniques [1], which perform a channel based noise reduction, temporal and spatial filtering [2,3]. Advanced beamforming also considers the correlation among different channels [4], and even extends the beamforming optimization with maximizing the speech recognition likelihood [5,6].…”

Signal processing methods for enhancing speech and music signals in reverberant environments

“…Thus, single or multi-channel blind deconvolution implementations usually involve very low channel orders and the number of reflections in the tested RIRs is unrealistically low (e.g. [42,52,77,78]). On the other hand, Miyoshi et al [120] have shown that in non-blind multichannel systems perfect inverse filtering can be achieved when the captured RIRs do not share any common zeros.…”