msvq

Functions

int main(int argc, char *argv[])

msvq [ option ] [ infile ]

• -l int

  • length of vector $(1\le M+1)$

• -m int

  • order of vector $(0\le M)$

• -s str

  • codebook file

• infile str

  • double-type vector to be quantized

• stdout

  • int-type codebook index

The below example quantizes and reconstructs vectors in data.d.

msvq -s cbfile < data.d | imsvq -s cbfile > data.q

Parameters:

• argc – [in] Number of arguments.

• argv – [in] Argument vector.

Returns:

0 on success, 1 on failure.

See also

imsvq lbg

class MultistageVectorQuantization

Perform multistage vector quantization.

The input is the $M$-th order vector:

$$\begin{array}{cccc}x(0),& x(1),& \dots ,& x(M),\end{array}$$

and the $M$-th order $N\times I$ codebook vectors, $\{c_i^{(n)}(m)\}$. The output is the $N$ codebook indices:

$$\begin{array}{cccc}i(1),& i(2),& \dots ,& i(N),\end{array}$$

where

$$i(n)=\underset{j}{\mathrm{argmin}}\sum _{j=0}^{I-1}\sum _{m=0}^M(e^{(n)}(m)-c_j^{(n)}(m){)}^2,$$

and the quantization error is

$$\begin{array}{r}{e}^{(n)}(m)=\{\begin{array}{ll}x(m),& n=1\\ e^{(n-1)}(m)-c_j^{(n-1)}(m).& n>1\end{array}\end{array}$$

Public Functions

MultistageVectorQuantization(int num_order, int num_stage)

Parameters:

• num_order – [in] Order of vector, $M$.

• num_stage – [in] Number of quantization stages, $N$.

inline int GetNumOrder() const

Returns:

Order of vector.

inline int GetNumStage() const

Returns:

Number of stages.

inline bool IsValid() const

Returns:

True if this object is valid.

bool Run(const std::vector<double> &input_vector, const std::vector<std::vector<std::vector<double>>> &codebook_vectors, std::vector<int> *codebook_indices, MultistageVectorQuantization::Buffer *buffer) const

Parameters:

• input_vector – [in] $M$-th order input vector.

• codebook_vectors – [in] $M$-th order $I$ codebook vectors. The shape is $[N,I,M+1]$.

• codebook_indices – [out] $N$ codebook indices.

• buffer – [out] Buffer.

Returns:

True on success, false on failure.

class Buffer

Buffer for MultistageVectorQuantization class.

class VectorQuantization

Perform vector quantization.

The input is the $M$-th order vector:

$$\begin{array}{cccc}x(0),& x(1),& \dots ,& x(M),\end{array}$$

and the $M$-th order codebook vectors:

$$\begin{array}{cccc}{\mathit{c}}_0,& {\mathit{c}}_1,& \dots ,& {\mathit{c}}_{I-1}.\end{array}$$

The output is the index of the codebook vector that minimizes the distance between the input vector and the codebook vector in an Euclidean sense:

$$\underset{i}{\mathrm{argmin}}\sum _{i=0}^{I-1}\sum _{m=0}^M(x(m)-c_i(m){)}^2.$$

Public Functions

explicit VectorQuantization(int num_order)

Parameters:

num_order – [in] Order of vector, $M$.

inline int GetNumOrder() const

Returns:

Order of vector.

inline bool IsValid() const

Returns:

True if this object is valid.

bool Run(const std::vector<double> &input_vector, const std::vector<std::vector<double>> &codebook_vectors, int *codebook_index) const

Parameters:

• input_vector – [in] $M$-th order input vector.

• codebook_vectors – [in] $M$-th order $I$ codebook vectors. The shape is $[I,M+1]$.

• codebook_index – [out] Codebook index.

Returns:

True on success, false on failure.