1// Copyright 2011 Dmitry Chestnykh. All rights reserved.
  2// Use of this source code is governed by a MIT-style
  3// license that can be found in the LICENSE file.
  4
  5package captcha
  6
  7import (
  8	"bytes"
  9	"encoding/binary"
 10	"math"
 11	"os"
 12	"rand"
 13	"io"
 14)
 15
 16const sampleRate = 8000 // Hz
 17
 18var (
 19	endingBeepSound []byte
 20)
 21
 22func init() {
 23	endingBeepSound = changeSpeed(beepSound, 1.4)
 24}
 25
 26// BUG(dchest): [Not our bug] Google Chrome 10 plays unsigned 8-bit PCM WAVE
 27// audio on Mac with horrible distortions.  Issue:
 28// http://code.google.com/p/chromium/issues/detail?id=70730.
 29// This has been fixed, and version 12 will play them properly.
 30
 31type Audio struct {
 32	body *bytes.Buffer
 33	digitSounds [][]byte
 34}
 35
 36// NewImage returns a new audio captcha with the given digits, where each digit
 37// must be in range 0-9. Digits are pronounced in the given language. If there
 38// are no sounds for the given language, English is used.
 39func NewAudio(digits []byte, lang string) *Audio {
 40	a := new(Audio)
 41	if sounds, ok := digitSounds[lang]; ok {
 42		a.digitSounds = sounds
 43	} else {
 44		a.digitSounds = digitSounds["en"]
 45	}
 46	numsnd := make([][]byte, len(digits))
 47	nsdur := 0
 48	for i, n := range digits {
 49		snd := a.randomizedDigitSound(n)
 50		nsdur += len(snd)
 51		numsnd[i] = snd
 52	}
 53	// Random intervals between digits (including beginning).
 54	intervals := make([]int, len(digits)+1)
 55	intdur := 0
 56	for i := range intervals {
 57		dur := rnd(sampleRate, sampleRate*3) // 1 to 3 seconds
 58		intdur += dur
 59		intervals[i] = dur
 60	}
 61	// Generate background sound.
 62	bg := a.makeBackgroundSound(a.longestDigitSndLen()*len(digits) + intdur)
 63	// Create buffer and write audio to it.
 64	sil := makeSilence(sampleRate / 5)
 65	bufcap := 3*len(beepSound) + 2*len(sil) + len(bg) + len(endingBeepSound)
 66	a.body = bytes.NewBuffer(make([]byte, 0, bufcap))
 67	// Write prelude, three beeps.
 68	a.body.Write(beepSound)
 69	a.body.Write(sil)
 70	a.body.Write(beepSound)
 71	a.body.Write(sil)
 72	a.body.Write(beepSound)
 73	// Write digits.
 74	pos := intervals[0]
 75	for i, v := range numsnd {
 76		mixSound(bg[pos:], v)
 77		pos += len(v) + intervals[i+1]
 78	}
 79	a.body.Write(bg)
 80	// Write ending (one beep).
 81	a.body.Write(endingBeepSound)
 82	return a
 83}
 84
 85// WriteTo writes captcha audio in WAVE format into the given io.Writer, and
 86// returns the number of bytes written and an error if any.
 87func (a *Audio) WriteTo(w io.Writer) (n int64, err os.Error) {
 88	// Calculate padded length of PCM chunk data.
 89	bodyLen := uint32(a.body.Len())
 90	paddedBodyLen := bodyLen
 91	if bodyLen%2 != 0 {
 92		paddedBodyLen++
 93	}
 94	totalLen := uint32(len(waveHeader)) - 4 + paddedBodyLen
 95	// Header.
 96	header := make([]byte, len(waveHeader)+4) // includes 4 bytes for chunk size
 97	copy(header, waveHeader)
 98	// Put the length of whole RIFF chunk.
 99	binary.LittleEndian.PutUint32(header[4:], totalLen)
100	// Put the length of WAVE chunk.
101	binary.LittleEndian.PutUint32(header[len(waveHeader):], bodyLen)
102	// Write header.
103	nn, err := w.Write(header)
104	n = int64(nn)
105	if err != nil {
106		return
107	}
108	// Write data.
109	n, err = a.body.WriteTo(w)
110	n += int64(nn)
111	if err != nil {
112		return
113	}
114	// Pad byte if chunk length is odd.
115	// (As header has even length, we can check if n is odd, not chunk).
116	if bodyLen != paddedBodyLen {
117		w.Write([]byte{0})
118		n++
119	}
120	return
121}
122
123// EncodedLen returns the length of WAV-encoded audio captcha.
124func (a *Audio) EncodedLen() int {
125	return len(waveHeader) + 4 + a.body.Len()
126}
127
128func (a *Audio) makeBackgroundSound(length int) []byte {
129	b := makeWhiteNoise(length, 4)
130	for i := 0; i < length/(sampleRate/10); i++ {
131		snd := reversedSound(a.digitSounds[rand.Intn(10)])
132		snd = changeSpeed(snd, rndf(0.8, 1.4))
133		place := rand.Intn(len(b) - len(snd))
134		setSoundLevel(snd, rndf(0.2, 0.5))
135		mixSound(b[place:], snd)
136	}
137	return b
138}
139
140func (a *Audio) randomizedDigitSound(n byte) []byte {
141	s := randomSpeed(a.digitSounds[n])
142	setSoundLevel(s, rndf(0.75, 1.2))
143	return s
144}
145
146func (a *Audio) longestDigitSndLen() int {
147	n := 0
148	for _, v := range a.digitSounds {
149		if n < len(v) {
150			n = len(v)
151		}
152	}
153	return n
154}
155
156// mixSound mixes src into dst. Dst must have length equal to or greater than
157// src length.
158func mixSound(dst, src []byte) {
159	for i, v := range src {
160		av := int(v)
161		bv := int(dst[i])
162		if av < 128 && bv < 128 {
163			dst[i] = byte(av * bv / 128)
164		} else {
165			dst[i] = byte(2*(av+bv) - av*bv/128 - 256)
166		}
167	}
168}
169
170func setSoundLevel(a []byte, level float64) {
171	for i, v := range a {
172		av := float64(v)
173		switch {
174		case av > 128:
175			if av = (av-128)*level + 128; av < 128 {
176				av = 128
177			}
178		case av < 128:
179			if av = 128 - (128-av)*level; av > 128 {
180				av = 128
181			}
182		default:
183			continue
184		}
185		a[i] = byte(av)
186	}
187}
188
189// changeSpeed returns new PCM bytes from the bytes with the speed and pitch
190// changed to the given value that must be in range [0, x].
191func changeSpeed(a []byte, speed float64) []byte {
192	b := make([]byte, int(math.Floor(float64(len(a))*speed)))
193	var p float64
194	for _, v := range a {
195		for i := int(p); i < int(p+speed); i++ {
196			b[i] = v
197		}
198		p += speed
199	}
200	return b
201}
202
203func randomSpeed(a []byte) []byte {
204	pitch := rndf(0.9, 1.2)
205	return changeSpeed(a, pitch)
206}
207
208func makeSilence(length int) []byte {
209	b := make([]byte, length)
210	for i := range b {
211		b[i] = 128
212	}
213	return b
214}
215
216func makeWhiteNoise(length int, level uint8) []byte {
217	noise := randomBytes(length)
218	adj := 128 - level/2
219	for i, v := range noise {
220		v %= level
221		v += adj
222		noise[i] = v
223	}
224	return noise
225}
226
227func reversedSound(a []byte) []byte {
228	n := len(a)
229	b := make([]byte, n)
230	for i, v := range a {
231		b[n-1-i] = v
232	}
233	return b
234}