-mfclogdir facility in PocketSphinx 0.5?

[Chia Tee-Kiah]

Hello all,

I have been using PocketSphinx 0.4.1 in a
project, and utilizing the -mfclogdir option to
record MFCCs of the heard utterances (for
purposes of speaker adaptation).

I am trying to do the same under PocketSphinx
0.5, but I could not find this feature in it.
What should I do if I wish to log the MFCCs
of utterances as they are input and recognized?
Are there any plans to include this facility
once more in later versions of PocketSphinx?

In case it is useful, the code I wrote to log
MFCCs goes roughly thus:

    static const char * const fake_argv[] = {
            ...
            "-mfclogdir",   MFCC_DIR, ...
    };
    ...
    fbs_init(argc, (char **)fake_argv);
    ...
    if (decode) {
            sprintf(utt_id, "%08x", utt_no);
            if (uttproc_begin_utt(utt_id) < 0) ...
            uttproc_rawdata(adbuf, k, 0);
    }
    ...
    for (...) {
            ...
            if (decode) {
                    rem = uttproc_rawdata(adbuf, k, 0);
                    ...
            }
    }
    ...
    if (decode) {
            uttproc_end_utt();
            if (uttproc_result(&fr, &hyp, 1) < 0) ...
            ...
    } ...

[David Huggins-Daines]

Hi! Sorry, it did get removed (or more precisely, it didn't get reimplemented) for 0.5. It is, in fact, fairly useful, so it would be nice to bring it back for the 0.5.1 point release (which will contain a variety of other bug fixes and missing features).

Can you file a bug?

Thanks!

[Chia Tee-Kiah]

I have done so, thanks!