Custom MP3 decoder from Scratch in Kotlin // Rascunho de decodificador de MP3 PRÓPRIO do ZERO

main.kt

@file:OptIn(ExperimentalUnsignedTypes::class)

import com.lucasalfare.flbinary.Reader
import com.lucasalfare.flbinary.readBits
import java.io.File

val bitrateTable = IntArray(0b1111) { -1 }
val sampleRateTable = IntArray(0b11) { -1 }

fun main() {
  bitrateTable[0b0001] = 32000
  bitrateTable[0b0010] = 40000
  bitrateTable[0b0011] = 48000
  bitrateTable[0b0100] = 56000
  bitrateTable[0b0101] = 64000
  bitrateTable[0b0110] = 80000
  bitrateTable[0b0111] = 96000
  bitrateTable[0b1000] = 112000
  bitrateTable[0b1001] = 128000
  bitrateTable[0b1010] = 160000
  bitrateTable[0b1011] = 192000
  bitrateTable[0b1100] = 224000
  bitrateTable[0b1101] = 256000
  bitrateTable[0b1110] = 320000

  sampleRateTable[0b00] = 44100
  sampleRateTable[0b01] = 48000
  sampleRateTable[0b10] = 32000

  val bytes = File("bip.mp3").readBytes().toUByteArray()
  val reader = Reader(bytes)
  skipTags(reader)
  println("Reading audio data...")

  var n = 0

  while (reader.position < bytes.size - 1) {
    val frameSynchronizer = reader.readBits(11)
    if (frameSynchronizer == 0b11111111_111L) {
      println("Found a frame header at position ${reader.position - 1}!")
      readAudioFrame(reader)
      println()
      n++
      if (n == 10) break
    } else {
      reader.position--
    }
  }
}

fun skipTags(reader: Reader) {
  val id3Signature = reader.readString(3)
  if (id3Signature == "ID3") {
    val versionMajor = reader.read1Byte()
    val versionMinor = reader.read1Byte()
    val flags = reader.read1Byte()

    val b3 = reader.read1Byte()
    val b2 = reader.read1Byte()
    val b1 = reader.read1Byte()
    val b0 = reader.read1Byte()

    val tagSize = ((b3 and 0x7F) shl 21) or
            ((b2 and 0x7F) shl 14) or
            ((b1 and 0x7F) shl 7) or
            (b0 and 0x7F)

    val hasExtendedHeader = (flags and 0x40) != 0
    var totalSize = tagSize + 10

    if (hasExtendedHeader) {
      val extHeaderSize = (reader.read1Byte() shl 24) or
              (reader.read1Byte() shl 16) or
              (reader.read1Byte() shl 8) or
              reader.read1Byte()

      val skipSize = if (versionMajor == 3) extHeaderSize - 4 else extHeaderSize
      reader.advancePosition(skipSize)
      totalSize += extHeaderSize
    }
    reader.advancePosition(totalSize - 10)
    println("Tag skipped ($totalSize bytes).")
  } else {
    reader.position -= 3
    println("No tags found.")
  }
}

fun readAudioFrame(reader: Reader) {
  reader.position -= 2

  // next bytes after frame synchronizer
  // we care only about the bytes 1/2/3 because byte 0 is filled with bits '1'.
  val headerBytes: UByteArray = reader.readBytes(4)
  val b2 = headerBytes[1].toInt()
  val b3 = headerBytes[2].toInt()
  val b4 = headerBytes[3].toInt()

  val mpegVersionID = (b2 shr 5) and 0b11
  val layer = (b2 shr 3) and 0b11
  val hasCrcProtection = (b2 shr 2) and 0b1

  val bitrateIndex = (b3 shr 4) and 0b1111
  val samplingRateFrequencyIndex = (b3 shr 2) and 0b11
  val hasPadding = (b3 shr 1) and 0b1
  val privateBit = b3 and 0b1

  val channelMode = (b4 shr 6) and 0b11
  val modeExtension = (b4 shr 4) and 0b11
  val isCopyrighted = (b4 shr 3) and 0b1
  val isOriginal = (b4 shr 2) and 0b1
  val emphasis = b4 and 0b11

  val bitrate = bitrateTable[bitrateIndex]
  val sampleRate = sampleRateTable[samplingRateFrequencyIndex]

  val currentFrameLength = (144 * bitrate / sampleRate) + hasPadding

//  println("[BB] modeExtension:  ${mpegVersionID.toString(2).padStart(2, '0')}")
//  println("[CC] layer:  ${layer.toString(2).padStart(2, '0')}")
//  println("[D] hasCrcProtection:   ${hasCrcProtection.toString(2).padStart(1, '0')}")
//  println("[EEEE] bitrateIndex:${bitrateIndex.toString(2).padStart(4, '0')}")
//  println("[FF] samplingRateFrequencyIndex:  ${samplingRateFrequencyIndex.toString(2).padStart(2, '0')}")
//  println("[G] hasPadding:   ${hasPadding.toString(2).padStart(1, '0')}")
//  println("[H] privateBit:   ${privateBit.toString(2).padStart(1, '0')}")
//  println("[II] channelMode:  ${channelMode.toString(2).padStart(2, '0')}")
//  println("[JJ] modeExtension:  ${modeExtension.toString(2).padStart(2, '0')}")
//  println("[K] isCopyrighted:   ${isCopyrighted.toString(2).padStart(1, '0')}")
//  println("[L] isOriginal:   ${isOriginal.toString(2).padStart(1, '0')}")
//  println("[MM] emphasis:  ${emphasis.toString(2).padStart(2, '0')}")
  println("Decoded bitrate: $bitrate")
  println("Decoded sampleRate: $sampleRate")
  println("Total frame length (in bytes): $currentFrameLength")
}