Reverberation removal device, parameter estimation device, reverberation removal method, parameter estimation method, and program
Provided is a reverberation removal device that is highly accurate even in noisy environments and underdetermined conditions. Reverberation is removed by applying a plurality of reverberation prediction filters to an observation signal while switching the plurality of reverberation prediction filters according to each time frequency bin of the observation signal.
1 . A reverberation removal device for removing reverberation, the device comprising a processor configured to execute operations comprising: storing the plurality of reverberation prediction filters; storing a mixed weight that determines switching to a first reverberation prediction filter of the reverberation prediction filters as one of the reverberation prediction filters to be applied according to each time frequency bin; and estimating a post-dereverberation signal z t in a time frame t by subtracting a result of computing the determined first reverberation prediction filter of the reverberation prediction filters predetermined by the mixed weight to an observation signal x − t in a predetermined section prior to the time frame t, from an observation signal x t in the time frame t and further by switching a second reverberation prediction filter of the plurality of reverberation prediction filters to the first reverberation prediction filter of the plurality of reverberation prediction filters to remove reverberation of audio.
2 . A computer-readable non-transitory recording medium storing a computer-executable program instructions that when executed by a processor cause a computer to execute as the reverberation removal device according to claim 1 .
3 . A parameter estimation device, comprising: processing circuitry configured to estimate a post-dereverberation signal z t in a time frame t by subtracting a result of computing a reverberation prediction filter of a plurality of reverberation prediction filters to an observation signal x − t in a predetermined section prior to the time frame t, from an observation signal x t in the time frame t; update (i) a mixed weight α t that determines which one of the reverberation prediction filters to be applied in accordance with each time frequency bin, and (ii) a power spectrum λ t obtained after dereverberation in the time frame t; update the reverberation prediction filters; and transmit a control command to repeatedly execute processing of the dereverberation, processing of the updating of both the mixed weight and the power spectrum steps, and processing of the reverberation prediction filter updating until an update amount of a parameter including the mixed weight α t , the power spectrum λ t and reverberation prediction filter becomes less than a predetermined threshold value.
4 . A computer-readable non-transitory recording medium storing a computer-executable program instructions that when executed by a processor cause a computer to execute as the parameter estimation device according to claim 3 .
5 . A reverberation removal method, comprising:
a post-dereverberation signal estimation step of estimating a post-dereverberation signal z t in a time frame t by subtracting a result of computing a reverberation prediction filter predetermined by a mixed weight to an observation signal
x
t
-
in a predetermined section prior to the time frame t, from an observation signal x t in the time frame t, by using the plurality of dereverberation prediction filters and a mixed weight determining switching to the reverberation prediction filter of the reverberation prediction filters to be applied in accordance with each time frequency bin and further by switching from another reverberation prediction filter to the reverberation prediction filter.
6 . A parameter estimation method, comprising: a dereverberation step of estimating a post-dereverberation signal z t in a time frame t by subtracting a result of computing any of a plurality of reverberation prediction filters to an observation signal
x
t
-
in a predetermined section prior to the time frame t, from an observation signal x t in the time frame t; a parameter updating step of updating a mixed weight α t that determines which one of the reverberation prediction filters should be applied in accordance with each time frequency bin, and a power spectrum λ t obtained after dereverberation in the time frame t; a reverberation prediction filter updating step of updating the reverberation prediction filters; and a control step of transmitting a control command to repeatedly execute processing of the dereverberation step, processing of the updating of both the mixed weight and the power spectrum step, and processing of the reverberation prediction filter updating step until an update amount of a parameter including the mixed weight α t , the power spectrum λ t and reverberation prediction filter becomes less than a predetermined threshold value.