Przemyslaw Szczepaniak | 90d15d2 | 2013-06-14 12:02:53 +0100 | [diff] [blame] | 1 | package android.speech.tts; |
| 2 | |
| 3 | import android.speech.tts.TextToSpeechClient.EngineStatus; |
| 4 | |
| 5 | import java.util.Locale; |
| 6 | |
| 7 | /** |
| 8 | * Set of common heuristics for selecting {@link VoiceInfo} from |
| 9 | * {@link TextToSpeechClient#getEngineStatus()} output. |
| 10 | */ |
| 11 | public final class RequestConfigHelper { |
| 12 | private RequestConfigHelper() {} |
| 13 | |
| 14 | /** |
| 15 | * Interface for scoring VoiceInfo object. |
| 16 | */ |
| 17 | public static interface VoiceScorer { |
| 18 | /** |
| 19 | * Score VoiceInfo. If the score is less than or equal to zero, that voice is discarded. |
| 20 | * If two voices have same desired primary characteristics (highest quality, lowest |
| 21 | * latency or others), the one with the higher score is selected. |
| 22 | */ |
| 23 | public int scoreVoice(VoiceInfo voiceInfo); |
| 24 | } |
| 25 | |
| 26 | /** |
| 27 | * Score positively voices that exactly match the locale supplied to the constructor. |
| 28 | */ |
| 29 | public static final class ExactLocaleMatcher implements VoiceScorer { |
| 30 | private final Locale mLocale; |
| 31 | |
| 32 | /** |
| 33 | * Score positively voices that exactly match the given locale |
| 34 | * @param locale Reference locale. If null, the default locale will be used. |
| 35 | */ |
| 36 | public ExactLocaleMatcher(Locale locale) { |
| 37 | if (locale == null) { |
| 38 | mLocale = Locale.getDefault(); |
| 39 | } else { |
| 40 | mLocale = locale; |
| 41 | } |
| 42 | } |
| 43 | @Override |
| 44 | public int scoreVoice(VoiceInfo voiceInfo) { |
| 45 | return mLocale.equals(voiceInfo.getLocale()) ? 1 : 0; |
| 46 | } |
| 47 | } |
| 48 | |
| 49 | /** |
| 50 | * Score positively voices that match exactly the given locale (score 3) |
| 51 | * or that share same language and country (score 2), or that share just a language (score 1). |
| 52 | */ |
| 53 | public static final class LanguageMatcher implements VoiceScorer { |
| 54 | private final Locale mLocale; |
| 55 | |
| 56 | /** |
| 57 | * Score positively voices with similar locale. |
| 58 | * @param locale Reference locale. If null, default will be used. |
| 59 | */ |
| 60 | public LanguageMatcher(Locale locale) { |
| 61 | if (locale == null) { |
| 62 | mLocale = Locale.getDefault(); |
| 63 | } else { |
| 64 | mLocale = locale; |
| 65 | } |
| 66 | } |
| 67 | |
| 68 | @Override |
| 69 | public int scoreVoice(VoiceInfo voiceInfo) { |
| 70 | final Locale voiceLocale = voiceInfo.getLocale(); |
| 71 | if (mLocale.equals(voiceLocale)) { |
| 72 | return 3; |
| 73 | } else { |
| 74 | if (mLocale.getLanguage().equals(voiceLocale.getLanguage())) { |
| 75 | if (mLocale.getCountry().equals(voiceLocale.getCountry())) { |
| 76 | return 2; |
| 77 | } |
| 78 | return 1; |
| 79 | } |
| 80 | return 0; |
| 81 | } |
| 82 | } |
| 83 | } |
| 84 | |
| 85 | /** |
| 86 | * Get the highest quality voice from voices that score more than zero from the passed scorer. |
| 87 | * If there is more than one voice with the same highest quality, then this method returns one |
| 88 | * with the highest score. If they share same score as well, one with the lower index in the |
| 89 | * voices list is returned. |
| 90 | * |
| 91 | * @param engineStatus |
| 92 | * Voices status received from a {@link TextToSpeechClient#getEngineStatus()} call. |
| 93 | * @param voiceScorer |
| 94 | * Used to discard unsuitable voices and help settle cases where more than |
| 95 | * one voice has the desired characteristic. |
| 96 | * @param hasToBeEmbedded |
| 97 | * If true, require the voice to be an embedded voice (no network |
| 98 | * access will be required for synthesis). |
| 99 | */ |
| 100 | private static VoiceInfo getHighestQualityVoice(EngineStatus engineStatus, |
| 101 | VoiceScorer voiceScorer, boolean hasToBeEmbedded) { |
| 102 | VoiceInfo bestVoice = null; |
| 103 | int bestScoreMatch = 1; |
| 104 | int bestVoiceQuality = 0; |
| 105 | |
| 106 | for (VoiceInfo voice : engineStatus.getVoices()) { |
| 107 | int score = voiceScorer.scoreVoice(voice); |
| 108 | if (score <= 0 || hasToBeEmbedded && voice.getRequiresNetworkConnection() |
| 109 | || voice.getQuality() < bestVoiceQuality) { |
| 110 | continue; |
| 111 | } |
| 112 | |
| 113 | if (bestVoice == null || |
| 114 | voice.getQuality() > bestVoiceQuality || |
| 115 | score > bestScoreMatch) { |
| 116 | bestVoice = voice; |
| 117 | bestScoreMatch = score; |
| 118 | bestVoiceQuality = voice.getQuality(); |
| 119 | } |
| 120 | } |
| 121 | return bestVoice; |
| 122 | } |
| 123 | |
| 124 | /** |
| 125 | * Get highest quality voice. |
| 126 | * |
| 127 | * Highest quality voice is selected from voices that score more than zero from the passed |
| 128 | * scorer. If there is more than one voice with the same highest quality, then this method |
| 129 | * will return one with the highest score. If they share same score as well, one with the lower |
| 130 | * index in the voices list is returned. |
| 131 | |
| 132 | * @param engineStatus |
| 133 | * Voices status received from a {@link TextToSpeechClient#getEngineStatus()} call. |
| 134 | * @param hasToBeEmbedded |
| 135 | * If true, require the voice to be an embedded voice (no network |
| 136 | * access will be required for synthesis). |
| 137 | * @param voiceScorer |
| 138 | * Scorer is used to discard unsuitable voices and help settle cases where more than |
| 139 | * one voice has highest quality. |
| 140 | * @return RequestConfig with selected voice or null if suitable voice was not found. |
| 141 | */ |
| 142 | public static RequestConfig highestQuality(EngineStatus engineStatus, |
| 143 | boolean hasToBeEmbedded, VoiceScorer voiceScorer) { |
| 144 | VoiceInfo voice = getHighestQualityVoice(engineStatus, voiceScorer, hasToBeEmbedded); |
| 145 | if (voice == null) { |
| 146 | return null; |
| 147 | } |
| 148 | return RequestConfig.Builder.newBuilder().setVoice(voice).build(); |
| 149 | } |
| 150 | |
| 151 | /** |
| 152 | * Get highest quality voice for the default locale. |
| 153 | * |
| 154 | * Call {@link #highestQuality(EngineStatus, boolean, VoiceScorer)} with |
| 155 | * {@link LanguageMatcher} set to device default locale. |
| 156 | * |
| 157 | * @param engineStatus |
| 158 | * Voices status received from a {@link TextToSpeechClient#getEngineStatus()} call. |
| 159 | * @param hasToBeEmbedded |
| 160 | * If true, require the voice to be an embedded voice (no network |
| 161 | * access will be required for synthesis). |
| 162 | * @return RequestConfig with selected voice or null if suitable voice was not found. |
| 163 | */ |
| 164 | public static RequestConfig highestQuality(EngineStatus engineStatus, |
| 165 | boolean hasToBeEmbedded) { |
| 166 | return highestQuality(engineStatus, hasToBeEmbedded, |
| 167 | new LanguageMatcher(Locale.getDefault())); |
| 168 | } |
| 169 | |
| 170 | } |