怎样在PHP项目中实现语音识别？

wen java案例 2026-07-21 172

本文目录导读：

怎样在PHP项目中实现语音识别？

使用云服务API（推荐方案）
使用WebSocket实时识别
使用本地语音识别引擎
音频预处理优化
完整的语音识别服务类
前端+后端完整流程
后端接收处理
注意事项

在PHP项目中实现语音识别,通常有以下几种主流方案：

使用云服务API（推荐方案）

百度语音识别

<?php
require_once 'AipSpeech.php';
// 初始化
$client = new AipSpeech('你的AppId', '你的API Key', '你的Secret Key');
// 读取音频文件
$audio = file_get_contents('./test.pcm');
// 识别
$result = $client->asr($audio, 'pcm', 16000, [
    'dev_pid' => 1536, // 普通话(支持简单的英文识别)
]);
if ($result['err_no'] == 0) {
    echo "识别结果：" . $result['result'];
} else {
    echo "错误：" . $result['err_msg'];
}
?>

阿里云语音识别

<?php
use AlibabaCloud\Client\AlibabaCloud;
use AlibabaCloud\Client\Exception\ClientException;
use AlibabaCloud\Client\Exception\ServerException;
// 配置
AlibabaCloud::accessKeyClient('your-access-key', 'your-access-secret')
    ->regionId('cn-shanghai')
    ->asDefaultClient();
try {
    $result = AlibabaCloud::nlsFiletrans()
        ->v20180817()
        ->createTask()
        ->withTask(json_encode([
            'appkey' => 'your-appkey',
            'file_link' => 'https://your-file-url/audio.wav',
            'version' => '4.0',
            'enable_words' => false
        ]))
        ->request();
    print_r($result->toArray());
} catch (ClientException $e) {
    echo $e->getErrorMessage() . PHP_EOL;
} catch (ServerException $e) {
    echo $e->getErrorMessage() . PHP_EOL;
}
?>

使用WebSocket实时识别

<?php
// 使用百度实时语音识别WebSocket
$ws = new WebSocket\Client("wss://ws-api.baidu.com/aip/auth");
$ws->send(json_encode([
    'type' => 'START',
    'appid' => 'your-appid',
    'token' => 'access-token',
    'params' => [
        'format' => 'pcm',
        'rate' => 16000,
        'dev_pid' => 1536,
    ]
]));
// 发送音频数据
$audioData = file_get_contents('audio.pcm');
$ws->send(bin2hex($audioData));
// 结束识别
$ws->send(json_encode(['type' => 'FINISH']));
// 接收结果
while (true) {
    $msg = $ws->receive();
    $data = json_decode($msg, true);
    if ($data['type'] == 'RESULT') {
        echo "识别结果：" . $data['result'] . "\n";
    }
    if ($data['type'] == 'FINISH') {
        break;
    }
}
?>

使用本地语音识别引擎

PocketSphinx（离线方案）

<?php
// 需要安装pocketsphinx-php扩展
$decoder = new PocketSphinx\Decoder();
$decoder->setConfiguration([
    'hmm' => '/path/to/acoustic/model',
    'lm' => '/path/to/language/model',
    'dict' => '/path/to/dictionary'
]);
$audio = file_get_contents('audio.wav');
$result = $decoder->decode($audio);
echo "识别结果：" . $result;
?>

音频预处理优化

音频格式转换

<?php
function convertAudio($inputFile) {
    // 使用FFmpeg转换音频格式
    $outputFile = tempnam(sys_get_temp_dir(), 'audio') . '.pcm';
    exec("ffmpeg -i {$inputFile} -ar 16000 -ac 1 -f s16le {$outputFile}", $output, $returnCode);
    if ($returnCode === 0) {
        return file_get_contents($outputFile);
    }
    return false;
}
?>

完整的语音识别服务类

<?php
class SpeechRecognitionService {
    private $provider; // baidu, aliyun, tencent
    public function __construct($provider = 'baidu') {
        $this->provider = $provider;
    }
    public function recognize($audioFile, $options = []) {
        switch ($this->provider) {
            case 'baidu':
                return $this->baiduRecognize($audioFile, $options);
            case 'aliyun':
                return $this->aliyunRecognize($audioFile, $options);
            default:
                throw new Exception("不支持的识别服务");
        }
    }
    private function baiduRecognize($audioFile, $options) {
        // 百度语音识别实现
        require_once 'AipSpeech.php';
        $client = new AipSpeech(
            getenv('BAIDU_APP_ID'),
            getenv('BAIDU_API_KEY'),
            getenv('BAIDU_SECRET_KEY')
        );
        // 自动转换音频格式
        $audioData = $this->prepareAudio($audioFile, 'pcm');
        return $client->asr($audioData, 'pcm', 16000, [
            'dev_pid' => $options['dev_pid'] ?? 1536,
        ]);
    }
    private function prepareAudio($audioFile, $format) {
        $info = pathinfo($audioFile);
        if ($info['extension'] !== 'pcm') {
            return $this->convertAudio($audioFile);
        }
        return file_get_contents($audioFile);
    }
}
?>

前端+后端完整流程

<!-- 前端HTML -->
<form action="upload.php" method="post" enctype="multipart/form-data">
    <input type="file" name="audio_file" accept="audio/*" capture>
    <button type="submit">上传并识别</button>
</form>
<!-- 或使用录音功能 -->
<script>
const recorder = new MediaRecorder(stream);
recorder.ondataavailable = async (event) => {
    const formData = new FormData();
    formData.append('audio', event.data);
    const response = await fetch('recognize.php', {
        method: 'POST',
        body: formData
    });
    const result = await response.json();
    console.log('识别结果:', result.text);
};
</script>

后端接收处理

// upload.php
<?php
if ($_FILES['audio']['error'] === UPLOAD_ERR_OK) {
    $tmpFile = $_FILES['audio']['tmp_name'];
    // 调用语音识别
    $speechService = new SpeechRecognitionService('baidu');
    $result = $speechService->recognize($tmpFile);
    // 保存到数据库
    $db->query("INSERT INTO speech_recognition SET 
        text = '{$result['result'][0]}',
        created_at = NOW()
    ");
    // 返回结果
    echo json_encode([
        'success' => true,
        'text' => $result['result'][0]
    ]);
}
?>