1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
|
//
// Copyright © 2020 Arm Ltd and Contributors. All rights reserved.
// SPDX-License-Identifier: MIT
//
#pragma once
#include "DataStructures.hpp"
#include "SlidingWindow.hpp"
#include <numeric>
#include "MFCC.hpp"
/* Class to facilitate pre-processing calculation for Wav2Letter model
* for ASR */
using AudioWindow = SlidingWindow <const float>;
class Preprocess
{
public:
MFCC _m_mfcc; /* MFCC instance */
/* Actual buffers to be populated */
Array2d<float> _m_mfccBuf; /* Contiguous buffer 1D: MFCC */
Array2d<float> _m_delta1Buf; /* Contiguous buffer 1D: Delta 1 */
Array2d<float> _m_delta2Buf; /* Contiguous buffer 1D: Delta 2 */
uint32_t _m_windowLen; /* Window length for MFCC */
uint32_t _m_windowStride; /* Window stride len for MFCC */
AudioWindow _m_window; /* Sliding window */
/**
* @brief Constructor
* @param[in] numMfccFeatures number of MFCC features per window
* @param[in] windowLen number of elements in a window
* @param[in] windowStride stride (in number of elements) for
* moving the window
* @param[in] numMfccVectors number of MFCC vectors per window
*/
Preprocess(
const uint32_t windowLen,
const uint32_t windowStride,
const MFCC mfccInst);
Preprocess() = delete;
~Preprocess();
/**
* @brief Calculates the features required from audio data. This
* includes MFCC, first and second order deltas,
* normalisation and finally, quantisation. The tensor is
* populated with feature from a given window placed along
* in a single row.
* @param[in] audioData pointer to the first element of audio data
* @param[in] audioDataLen number of elements in the audio data
* @param[in] tensor tensor to be populated
* @return true if successful, false in case of error.
*/
bool Invoke(const float* audioData,
const uint32_t audioDataLen,
std::vector<int8_t>& output,
int quantOffset,
float quantScale);
protected:
/**
* @brief Computes the first and second order deltas for the
* MFCC buffers - they are assumed to be populated.
*
* @param[in] mfcc MFCC buffers
* @param[out] delta1 result of the first diff computation
* @param[out] delta2 result of the second diff computation
*
* @return true if successful, false otherwise
*/
static bool _ComputeDeltas(Array2d<float>& mfcc,
Array2d<float>& delta1,
Array2d<float>& delta2);
/**
* @brief Given a 2D vector of floats, computes the mean
* @param[in] vec vector of vector of floats
* @return mean value
*/
static float _GetMean(Array2d<float>& vec);
/**
* @brief Given a 2D vector of floats, computes the stddev
* @param[in] vec vector of vector of floats
* @param[in] mean mean value of the vector passed in
* @return stddev value
*/
static float _GetStdDev(Array2d<float>& vec,
const float mean);
/**
* @brief Given a 2D vector of floats, normalises it using
* the mean and the stddev
* @param[in/out] vec vector of vector of floats
* @return
*/
static void _NormaliseVec(Array2d<float>& vec);
/**
* @brief Normalises the MFCC and delta buffers
* @return
*/
void _Normalise();
/**
* @brief Given the quantisation and data type limits, computes
* the quantised values of a floating point input data.
* @param[in] elem Element to be quantised
* @param[in] quantScale Scale
* @param[in] quantOffset Offset
* @param[in] minVal Numerical limit - minimum
* @param[in] maxVal Numerical limit - maximum
* @return floating point quantised value
*/
static float _GetQuantElem(
const float elem,
const float quantScale,
const int quantOffset,
const float minVal,
const float maxVal);
/**
* @brief Quantises the MFCC and delta buffers, and places them
* in the output buffer. While doing so, it transposes
* the data. Reason: Buffers in this class are arranged
* for "time" axis to be row major. Primary reason for
* this being the convolution speed up (as we can use
* contiguous memory). The output, however, requires the
* time axis to be in column major arrangement.
* @param[in] outputBuf pointer to the output buffer
* @param[in] outputBufSz output buffer's size
* @param[in] quantScale quantisation scale
* @param[in] quantOffset quantisation offset
*/
template <typename T>
bool _Quantise(T* outputBuf, int quantOffset, float quantScale)
{
/* Populate */
T* outputBufMfcc = outputBuf;
T* outputBufD1 = outputBuf + this->_m_mfcc._m_params.m_numMfccFeatures;
T* outputBufD2 = outputBufD1 + this->_m_mfcc._m_params.m_numMfccFeatures;
const uint32_t ptrIncr = this->_m_mfcc._m_params.m_numMfccFeatures * 2; /* (3 vectors - 1 vector) */
const float minVal = std::numeric_limits<T>::min();
const float maxVal = std::numeric_limits<T>::max();
/* We need to do a transpose while copying and concatenating
* the tensor*/
for (uint32_t j = 0; j < this->_m_mfcc._m_params.m_numMfccVectors; ++j) {
for (uint32_t i = 0; i < this->_m_mfcc._m_params.m_numMfccFeatures; ++i)
{
*outputBufMfcc++ = static_cast<T>(this->_GetQuantElem(
this->_m_mfccBuf(i, j), quantScale,
quantOffset, minVal, maxVal));
*outputBufD1++ = static_cast<T>(this->_GetQuantElem(
this->_m_delta1Buf(i, j), quantScale,
quantOffset, minVal, maxVal));
*outputBufD2++ = static_cast<T>(this->_GetQuantElem(
this->_m_delta2Buf(i, j), quantScale,
quantOffset, minVal, maxVal));
}
outputBufMfcc += ptrIncr;
outputBufD1 += ptrIncr;
outputBufD2 += ptrIncr;
}
return true;
}
};
|