search for: compute_rm

...ex_bits_read_from speex_bits_flush speex_bits_read_whole_bytes speex_bits_write speex_bits_write_whole_bytes speex_bits_unpack_signed speex_bits_peek_unsigned speex_bits_peek speex_bits_nbytes speex_bits_insert_terminator In cd_search.c: noise_codebook_quant noise_codebook_unquant In filters.c: compute_rms syn_percep_zero16 qmf_decomp qmf_synth In ltp.c: forced_pitch_quant forced_pitch_unquant In math_approx.c: spx_ilog2 _spx_cos_pi_2 spx_cos_norm spx_exp2 spx_atan In nb_celp.c: nb_encoder_destroy nb_decoder_destroy In quant_lsp.c: lsp_quant_nb lsp_unquant_nb lsp_quant_high lsp_unquant_high...

..., but I'm surprised you're not at least using speex_bits_init(), speex_bits_read_from() and speex_bits_write(). What are you using instead? > In cd_search.c: > noise_codebook_quant > noise_codebook_unquant No need for that (assuming you only want 8 kbps). > In filters.c: > compute_rms > syn_percep_zero16 > qmf_decomp > qmf_synth No need for that. > In ltp.c: > forced_pitch_quant > forced_pitch_unquant No need for that. > In math_approx.c: > spx_ilog2 > _spx_cos_pi_2 > spx_cos_norm > spx_exp2 > spx_atan No need for that. > In nb_celp....

...eex_bits_write > speex_bits_write_whole_bytes > speex_bits_unpack_signed > speex_bits_peek_unsigned > speex_bits_peek > speex_bits_nbytes > speex_bits_insert_terminator > > In cd_search.c: > noise_codebook_quant > noise_codebook_unquant > > In filters.c: > compute_rms > syn_percep_zero16 > qmf_decomp > qmf_synth > > In ltp.c: > forced_pitch_quant > forced_pitch_unquant > > In math_approx.c: > spx_ilog2 > _spx_cos_pi_2 > spx_cos_norm > spx_exp2 > spx_atan > > In nb_celp.c: > nb_encoder_destroy > nb_deco...

...utput signal entirly > zero(like a silenze). > When I was debugging I dicovered that there are a lot > of multiplications that produce floating > point overflow exceptions(Not A Number). > functions like compute_rms(const spx_sig_t *x, int > len) in file filters.c produce this kind of > behaviour. > I don't know if you alredy know this "problem" or not, > but, if you want I can send you in a raw file >...

...>> speex_bits_write speex_bits_write_whole_bytes >> speex_bits_unpack_signed speex_bits_peek_unsigned speex_bits_peek >> speex_bits_nbytes speex_bits_insert_terminator >> >> In cd_search.c: noise_codebook_quant noise_codebook_unquant >> >> In filters.c: compute_rms syn_percep_zero16 qmf_decomp qmf_synth >> >> In ltp.c: forced_pitch_quant forced_pitch_unquant >> >> In math_approx.c: spx_ilog2 _spx_cos_pi_2 spx_cos_norm spx_exp2 >> spx_atan >> >> In nb_celp.c: nb_encoder_destroy nb_decoder_destroy >> >&gt...

...eal" excitation*/ fir_mem2(st->frame, st->interp_lpc, st->exc, st->frameSize, st->lpcSize, st->mem_exc); /* Compute open-loop excitation gain */ #ifdef EPIC_48K if (st->lbr_48k) { float ol1=0,ol2=0; float ol_gain2; ol1 = compute_rms(st->exc, st->frameSize>>1); ol2 = compute_rms(st->exc+(st->frameSize>>1), st->frameSize>>1); ol1 *= ol1*(st->frameSize>>1); ol2 *= ol2*(st->frameSize>>1); ol_gain2=ol1; if (ol2>ol1) ol_g...

...tmp = NEG32(tmp); if (tmp >= max_val) max_val = tmp; } It should really use tmp=NEG32(tmp+1), since the range of negative fixed point has one more number for a given number of bits. (and that's what all the shifting is about) This same code fragment is in normalize16() and compute_rms() and a couple of other spots. I think it should be encapsulated into a norm_shift() function that can be overridden for specific processors. Similarly, the AEC code uses FLOAT_ADD, _SUB, _MULT routines in pseudofloat.h could be improved by shifting negative mantissa's >= -16384. if (...

You might not be doing anything wrong. Speex is stateful and can get into a bad state sometimes. It's happened to me too, but not recently. And I've seen such bugs reported on this list and fixed. I suggest several approaches to solving the problem: 1) Try compiling Speex without compiler optimizations and see if the problem still occurs. Most likely it will, but this is an

...l1_bound -R exc_gain_quant_scal3 -R exc_gain_quant_scal3_bound -R gain_cdbk_lbr -R gain_cdbk_nb -R hexc_10_32_table -R hexc_table -R high_lsp_cdbk -R high_lsp_cdbk2 -R lag_window -R lpc_window -R shift_filt -R vbr_hb_thresh -R vbr_nb_thresh -R vbr_uhb_thresh -T bw_lpc -T compute_impulse_response -T compute_rms -T compute_rms16 -T filter_mem16 T _fini -T fir_mem16 -T forced_pitch_quant -T forced_pitch_unquant -T highpass -T iir_mem16 T _init -T inner_prod -T interp_pitch -T lpc_to_lsp -T lsp_enforce_margin -T lsp_interpolate -T lsp_quant_high -T lsp_quant_lbr -T lsp_quant_nb -T lsp_to_lpc -T lsp_unquan...

...l1_bound -R exc_gain_quant_scal3 -R exc_gain_quant_scal3_bound -R gain_cdbk_lbr -R gain_cdbk_nb -R hexc_10_32_table -R hexc_table -R high_lsp_cdbk -R high_lsp_cdbk2 -R lag_window -R lpc_window -R shift_filt -R vbr_hb_thresh -R vbr_nb_thresh -R vbr_uhb_thresh -T bw_lpc -T compute_impulse_response -T compute_rms -T compute_rms16 -T filter_mem16 T _fini -T fir_mem16 -T forced_pitch_quant -T forced_pitch_unquant -T highpass -T iir_mem16 T _init -T inner_prod -T interp_pitch -T lpc_to_lsp -T lsp_enforce_margin -T lsp_interpolate -T lsp_quant_high -T lsp_quant_lbr -T lsp_quant_nb -T lsp_to_lpc -T lsp_unquan...