Method and apparatus for improving speech recognition processing performance

An apparatus configured to perform speech recognition processing, the apparatus comprising:

• a processor configured to:
• produce at least one first high dimensional feature vector by transforming at least one input feature vector produced from at least one speech signal into a high dimensional space;
• produce at least one second high dimensional feature vector by multiplying the at least one first high dimensional feature vector with a matrix, the multiplying including (i) obtaining a table value, from a plurality of table values each corresponding to a unique summation of vector element values of a unique combination of vector elements of a first vector, via an index corresponding to an encoded form of a combination of quantized element values of a second vector, the first vector based on the at least one first high dimensional feature vector and the second vector based on the matrix and (ii) computing a dot product value of the first and second vectors using the table value obtained, the vector element values, and the quantized element values, wherein the computing causes the processor to use fewer mathematical operations to compute the dot product value relative to a standard dot product computation of the first and second vectors to improve a computation time of the speech processing, the speech recognition processing performing multiple dot product computations to produce the at least one second high dimensional feature vector;
• and
• produce at least one output feature vector based on the at least one second high dimensional feature vector and employ the at least one output feature vector in conjunction with an acoustic model library to produce at least one speech recognition result based on the at least one speech signal.

A non-transitory computer-readable medium for improving computation time of speech recognition processing in an electronic device, the non-transitory computer-readable medium having encoded thereon a sequence of instructions which, when executed by a processor, causes the processor to:

• produce at least one first high dimensional feature vector by transforming at least one input feature vector produced from at least one speech signal into a high dimensional space;
• produce at least one second high dimensional feature vector by multiplying the at least one first high dimensional feature vector with a matrix, the multiplying including (i) obtaining a table value, from a plurality of table values each corresponding to a unique summation of vector element values of a unique combination of vector elements of a first vector, via an index corresponding to an encoded form of a combination of quantized element values of a second vector, the first vector based on the at least one first high dimensional feature vector and the second vector based on the matrix, and (ii) computing a dot product value of the first and second vectors using the table value obtained, the vector element values, and the quantized element values, wherein the computing causes the processor to use fewer mathematical operations to compute the dot product value relative to a standard dot product computation of the first and second vector to improve the computation time of the speech processing, the processor performing multiple dot product computations to produce the at least one second high dimensional feature vector;
• and
• produce at least one output feature vector based on the at least one second high dimensional feature vector and employ the at least one output feature vector in conjunction with an acoustic model library to produce at least one speech recognition result based on the at least one speech signal.