MFCC F16

RISCV_DSP_ATTRIBUTE void riscv_mfcc_f16 (const riscv_mfcc_instance_f16 *S, float16_t *pSrc, float16_t *pDst, float16_t *pTmp)
RISCV_DSP_ATTRIBUTE riscv_status riscv_mfcc_init_f16 (riscv_mfcc_instance_f16 *S, uint32_t fftLen, uint32_t nbMelFilters, uint32_t nbDctOutputs, const float16_t *dctCoefs, const uint32_t *filterPos, const uint32_t *filterLengths, const float16_t *filterCoefs, const float16_t *windowCoefs)
MFCC_INIT_F16(LEN)
group MFCCF16

Defines

MFCC_INIT_F16(LEN)

Functions

RISCV_DSP_ATTRIBUTE void riscv_mfcc_f16 (const riscv_mfcc_instance_f16 *S, float16_t *pSrc, float16_t *pDst, float16_t *pTmp)

MFCC F16.

The temporary buffer has a 2*fft length size when MFCC is implemented with CFFT. It has length FFT Length + 2 when implemented with RFFT (default implementation).

Description

The number of input samples if the FFT length used when initializing the instance data structure.

The source buffer is modified by this function.

Parameters

  • S – [in] points to the mfcc instance structure

  • pSrc – [in] points to the input samples

  • pDst – [out] points to the output MFCC values

  • pTmp – [inout] points to a temporary buffer of complex

RISCV_DSP_ATTRIBUTE riscv_status riscv_mfcc_init_f16 (riscv_mfcc_instance_f16 *S, uint32_t fftLen, uint32_t nbMelFilters, uint32_t nbDctOutputs, const float16_t *dctCoefs, const uint32_t *filterPos, const uint32_t *filterLengths, const float16_t *filterCoefs, const float16_t *windowCoefs)

Generic initialization of the MFCC F16 instance structure.

window coefficients can describe (for instance) a Hamming window. The array has the same size as the FFT length.

Description

The matrix of Mel filter coefficients is sparse. Most of the coefficients are zero. To avoid multiplying the spectrogram by those zeros, the filter is applied only to a given position in the spectrogram and on a given number of FFT bins (the filter length). It is the reason for the arrays filterPos and filterLengths.

The folder Scripts is containing a Python script that can be used to generate the filter, dct and window arrays.

This function should be used only if you don’t know the FFT sizes that you’ll need at build time. The use of this function will prevent the linker from removing the FFT tables that are not needed and the library code size will be bigger than needed.

If you use NMSIS-DSP as a static library, and if you know the MFCC sizes that you need at build time, then it is better to use the initialization functions defined for each MFCC size.

Parameters

  • S – [out] points to the mfcc instance structure

  • fftLen – [in] fft length

  • nbMelFilters – [in] number of Mel filters

  • nbDctOutputs – [in] number of Dct outputs

  • dctCoefs – [in] points to an array of DCT coefficients

  • filterPos – [in] points of the array of filter positions

  • filterLengths – [in] points to the array of filter lengths

  • filterCoefs – [in] points to the array of filter coefficients

  • windowCoefs – [in] points to the array of window coefficients

Returns

error status