audio.go 4.5 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205
  1. package captcha
  2. import (
  3. "bytes"
  4. "encoding/binary"
  5. "math"
  6. "os"
  7. "rand"
  8. "io"
  9. )
  10. const sampleRate = 8000 // Hz
  11. var (
  12. longestDigitSndLen int
  13. endingBeepSound []byte
  14. reverseDigitSounds [][]byte
  15. )
  16. func init() {
  17. for _, v := range digitSounds {
  18. if longestDigitSndLen < len(v) {
  19. longestDigitSndLen = len(v)
  20. }
  21. }
  22. endingBeepSound = changeSpeed(beepSound, 1.4)
  23. // Preallocate reversed digit sounds for background noise.
  24. reverseDigitSounds = make([][]byte, len(digitSounds))
  25. for i, v := range digitSounds {
  26. reverseDigitSounds[i] = reversedSound(v)
  27. }
  28. }
  29. // BUG(dchest): [Not our bug] Google Chrome 10 plays unsigned 8-bit PCM WAVE
  30. // audio on Mac with horrible distortions. Issue:
  31. // http://code.google.com/p/chromium/issues/detail?id=70730.
  32. // This has been fixed, and version 12 will play them properly.
  33. type Audio struct {
  34. body *bytes.Buffer
  35. }
  36. // NewImage returns a new audio captcha with the given digits, where each digit
  37. // must be in range 0-9.
  38. func NewAudio(digits []byte) *Audio {
  39. numsnd := make([][]byte, len(digits))
  40. nsdur := 0
  41. for i, n := range digits {
  42. snd := randomizedDigitSound(n)
  43. nsdur += len(snd)
  44. numsnd[i] = snd
  45. }
  46. // Random intervals between digits (including beginning).
  47. intervals := make([]int, len(digits)+1)
  48. intdur := 0
  49. for i := range intervals {
  50. dur := rnd(sampleRate, sampleRate*3) // 1 to 3 seconds
  51. intdur += dur
  52. intervals[i] = dur
  53. }
  54. // Generate background sound.
  55. bg := makeBackgroundSound(longestDigitSndLen*len(digits) + intdur)
  56. // Create buffer and write audio to it.
  57. a := new(Audio)
  58. sil := makeSilence(sampleRate / 5)
  59. bufcap := 3*len(beepSound) + 2*len(sil) + len(bg) + len(endingBeepSound)
  60. a.body = bytes.NewBuffer(make([]byte, 0, bufcap))
  61. // Write prelude, three beeps.
  62. a.body.Write(beepSound)
  63. a.body.Write(sil)
  64. a.body.Write(beepSound)
  65. a.body.Write(sil)
  66. a.body.Write(beepSound)
  67. // Write digits.
  68. pos := intervals[0]
  69. for i, v := range numsnd {
  70. mixSound(bg[pos:], v)
  71. pos += len(v) + intervals[i+1]
  72. }
  73. a.body.Write(bg)
  74. // Write ending (one beep).
  75. a.body.Write(endingBeepSound)
  76. return a
  77. }
  78. // WriteTo writes captcha audio in WAVE format into the given io.Writer, and
  79. // returns the number of bytes written and an error if any.
  80. func (a *Audio) WriteTo(w io.Writer) (n int64, err os.Error) {
  81. nn, err := w.Write(waveHeader)
  82. n = int64(nn)
  83. if err != nil {
  84. return
  85. }
  86. err = binary.Write(w, binary.LittleEndian, uint32(a.body.Len()))
  87. if err != nil {
  88. return
  89. }
  90. nn += 4
  91. n, err = a.body.WriteTo(w)
  92. n += int64(nn)
  93. return
  94. }
  95. // EncodedLen returns the length of WAV-encoded audio captcha.
  96. func (a *Audio) EncodedLen() int {
  97. return len(waveHeader) + 4 + a.body.Len()
  98. }
  99. // mixSound mixes src into dst. Dst must have length equal to or greater than
  100. // src length.
  101. func mixSound(dst, src []byte) {
  102. for i, v := range src {
  103. av := int(v)
  104. bv := int(dst[i])
  105. if av < 128 && bv < 128 {
  106. dst[i] = byte(av * bv / 128)
  107. } else {
  108. dst[i] = byte(2*(av+bv) - av*bv/128 - 256)
  109. }
  110. }
  111. }
  112. func setSoundLevel(a []byte, level float64) {
  113. for i, v := range a {
  114. av := float64(v)
  115. switch {
  116. case av > 128:
  117. if av = (av-128)*level + 128; av < 128 {
  118. av = 128
  119. }
  120. case av < 128:
  121. if av = 128 - (128-av)*level; av > 128 {
  122. av = 128
  123. }
  124. default:
  125. continue
  126. }
  127. a[i] = byte(av)
  128. }
  129. }
  130. // changeSpeed returns new PCM bytes from the bytes with the speed and pitch
  131. // changed to the given value that must be in range [0, x].
  132. func changeSpeed(a []byte, speed float64) []byte {
  133. b := make([]byte, int(math.Floor(float64(len(a))*speed)))
  134. var p float64
  135. for _, v := range a {
  136. for i := int(p); i < int(p+speed); i++ {
  137. b[i] = v
  138. }
  139. p += speed
  140. }
  141. return b
  142. }
  143. func randomSpeed(a []byte) []byte {
  144. pitch := rndf(0.9, 1.2)
  145. return changeSpeed(a, pitch)
  146. }
  147. func makeSilence(length int) []byte {
  148. b := make([]byte, length)
  149. for i := range b {
  150. b[i] = 128
  151. }
  152. return b
  153. }
  154. func makeWhiteNoise(length int, level uint8) []byte {
  155. noise := randomBytes(length)
  156. adj := 128 - level/2
  157. for i, v := range noise {
  158. v %= level
  159. v += adj
  160. noise[i] = v
  161. }
  162. return noise
  163. }
  164. func reversedSound(a []byte) []byte {
  165. n := len(a)
  166. b := make([]byte, n)
  167. for i, v := range a {
  168. b[n-1-i] = v
  169. }
  170. return b
  171. }
  172. func makeBackgroundSound(length int) []byte {
  173. b := makeWhiteNoise(length, 4)
  174. for i := 0; i < length/(sampleRate/10); i++ {
  175. snd := reverseDigitSounds[rand.Intn(10)]
  176. snd = changeSpeed(snd, rndf(0.8, 1.4))
  177. place := rand.Intn(len(b) - len(snd))
  178. setSoundLevel(snd, rndf(0.2, 0.3))
  179. mixSound(b[place:], snd)
  180. }
  181. return b
  182. }
  183. func randomizedDigitSound(n byte) []byte {
  184. s := randomSpeed(digitSounds[n])
  185. setSoundLevel(s, rndf(0.7, 1.3))
  186. return s
  187. }